# Nishanthan Janarthanarajah (Nishy): founder of Agent Studio and Flows > Founder of Agent Studio and Flows · Associate Technical Lead at Eight25Media · Colombo, Sri Lanka Independent AI engineer building AI agents, RAG assistants and workflow automation for businesses in Sri Lanka and worldwide, and creator of two live AI products, Flows and Agent Studio. **Availability:** Accepting new client projects: custom AI agents, workflow automation, RAG assistants, AI features and MCP servers. Remote-first from Colombo, Sri Lanka, with overlap for US, UK, European and Asia-Pacific hours. Replies within a day. Full Stack Developer and Engineering Lead with 7+ years architecting and delivering scalable, high-performance solutions across JavaScript/TypeScript, Next.js, React, Node.js, headless CMS, and cloud platforms. I lead cross-functional engineering teams end-to-end, from architecture through production, while aligning technical execution with business and marketing goals. I've scaled engineering teams from 6 to 31 engineers, directed enterprise B2B platform delivery, and built trusted client relationships, with deep experience in Next.js/Contentful/Vercel architecture, technical SEO, caching and performance engineering, CI/CD pipelines, and AWS. I'm increasingly focused on applied Generative AI and LLM application engineering: designing production Retrieval-Augmented Generation (RAG) systems on Azure AI Foundry and Azure OpenAI with Pinecone vector search and role-aware prompt guardrails, and building Model Context Protocol (MCP) servers that let AI agents drive design-to-code generation and headless CMS content automation. If you want to put AI agents to work in your business — agentic workflows, RAG assistants over your own data, MCP servers, or AI features inside your product — I'd love to talk about what you're building. ## Machine-readable surfaces - https://www.nishyai.com/for-agents (text/html): Agent view: every key fact on one HTML page with full JSON-LD - https://www.nishyai.com/llms.txt (text/markdown): Short summary for LLMs - https://www.nishyai.com/llms-full.txt (text/markdown): Full site content as markdown - https://www.nishyai.com/agent.json (application/json): Structured JSON of the same facts - https://www.nishyai.com/sitemap.xml (application/xml): Sitemap Every marketing page also has a markdown twin: request it with `Accept: text/markdown` or open the /md/ path listed under Pages. ## Products ### Flows: Describe the process. We run it. Describe a business process in plain English. Flows connects your apps, builds the workflow, and runs it for you. Every flow, run and log is stored in Neon serverless Postgres, so runs survive restarts and history is queryable. - Product: https://flows.nishyai.com/ - Landing page: https://www.nishyai.com/flows - Stack: Next.js, Clerk, Claude API, Neon Postgres, Tailwind CSS - Pricing: Free $0/month · Starter $10/month · Growth $49/month · Agency $200/month - Compared with: Zapier, Make, n8n - Website lead → alert you, thank them: The lead is in your inbox and Slack within seconds, and the prospect gets an instant thank-you written in your voice. - Claude reads each enquiry and flags the hot ones: Every enquiry is scored hot, warm or cold. Only hot leads interrupt you, with the reason and a suggested reply. - Daily morning digest, written for you: Every weekday at 8am, pull yesterday's numbers from any API and get a five-line briefing in your inbox. - Failed payment → recover it politely: A friendly retry link goes out straight away, and you get a reminder a day later if the invoice is still open. - One playbook, every client: Set up onboarding, reporting and lead routing once, then reuse it across client workspaces on the Agency plan. - Triage the support inbox: New tickets are summarised, tagged by urgency and routed to the right person, so nobody reads every email twice. ### Agent Studio: Design the agent. We run it as an API. Design an AI agent on a canvas, wrap it in guardrails and a prompt-injection shield, and deploy it as an API. Every design compiles into an immutable, versioned deployment behind one stable endpoint, and every call is logged with its stage-by-stage trace, latency and token usage in Neon serverless Postgres. - Product: https://agents.nishyai.com/ - Landing page: https://www.nishyai.com/agent-studio - Stack: Next.js, Groq · GPT-OSS & Llama, Llama Prompt Guard 2, Neon Postgres, Polar, Vercel - Pricing: Free trial $0/month · Starter $10/month · Pro $25/month · Growth $49/month · Agency $200/month - Compared with: OpenAI Agent Builder, Microsoft Copilot Studio, Dify, Flowise / Langflow - Support copilot with an order tracker: The orchestrator answers customers; an Order Tracker sub-agent looks up shipping status. A guardrail keeps it on topic and the shield blocks jailbreak attempts. - Internal knowledge assistant: Answers policy and process questions for staff. An output guardrail rewrites anything that leaks confidential detail, and every answer carries its trace. - Lead qualification agent behind a form: Call the agent from your website form. It asks two follow-up questions, scores the lead and returns structured output your CRM can store. - Brand-safe marketing assistant: Drafts posts and replies inside a natural-language brand policy. Off-brand output is rewritten to comply rather than blocked. - Triage agent that routes to specialists: One orchestrator reads the request and hands it to the right specialist: billing, technical, or account. Each specialist has its own model and prompt. - Twenty client agents on one account: The Agency plan runs up to 20 agents with 20 sub-agents each. Every client gets its own endpoint, API key and version history. ## Custom services - **Custom AI agents that do real work**: Agents designed around your process: they plan, call your tools and finish the job, from intake and research to drafting and updating your systems, with approval gates so autonomous work stays safe. - **RAG assistants over your data**: Assistants that answer from your documents, policies and databases with citations. Built in production on Azure AI Foundry and Azure OpenAI with Pinecone vector search and role-aware guardrails. - **Custom workflow automation**: Lead capture, reporting, onboarding, support triage or whatever your team repeats every day, wired across the tools you already use. Delivered on Flows when it fits, or built bespoke with the integrations you need. - **AI features inside your product**: Copilots, chat and generation features embedded in your Next.js or Node.js app, with prompt engineering, evaluation and content safety built in. - **MCP servers and agent tooling**: Model Context Protocol servers that expose your CMS, design tools, APIs and codebase to AI agents. Proven on Figma design-to-code and headless CMS content automation. - **AI discovery and proof of concept**: A short, fixed-scope engagement to find where AI actually pays off in your business, a working prototype, and a roadmap to production. ## Production proof - **Enterprise RAG HR assistant**: Retrieval-Augmented Generation assistant on Azure AI Foundry with Pinecone vector search, re-ranking and role-based prompt guardrails, answering staff questions over internal policy. - **AI interview and screening platform**: Candidate screening with LLM-as-a-judge scoring, structured rubrics and human review, built for a hiring team that needed consistency at volume. - **PandaDoc marketing platform (US)**: Lead engineering for PandaDoc's B2B site on Next.js, Contentful and Vercel at Eight25media: SEO strategy, a Contentful–Smartling localization pipeline and dual-layer caching. - **Paycor data migration and Modjoul IoT (US)**: Led the Snowflake migration of Paycor's on-prem client data, and architected secure AWS API Gateway and Cognito integrations for Modjoul's IoT asset-management platform. - **Flows and Agent Studio**: Two AI products designed, built and operated end to end: Next.js on Vercel, Neon serverless Postgres, Claude and Groq models, with paying plans and free tiers. - **Scaling a delivery team**: Promoted from Senior Engineer to Associate Technical Lead at Verdentra, directing staff augmentation for Paycor and growing the team from 6 to 31 engineers. ## How an engagement works 1. **Talk**: A WhatsApp, email or video call about the job you want done. I tell you honestly whether AI is the right tool, and whether Flows or Agent Studio already covers it. 2. **Discover**: A short, fixed-scope discovery: your data, your tools, the risks, and a prototype you can click. 3. **Build**: Production build with evaluation, guardrails and monitoring, delivered in weekly increments you can see. 4. **Run**: Handover with documentation, or I keep operating it for you. Either way you own the code and the data. ## Skills JavaScript, TypeScript, Python, .NET, Next.js, React, Node.js, NestJS, GraphQL, REST APIs, Contentful, Contentstack, Headless CMS, Vercel, AWS, Azure AI Foundry, Azure OpenAI, RAG, Pinecone, Vector Search, MCP Servers, AI Agents, Prompt Engineering, LLM Evaluation, Snowflake, MySQL, MongoDB, Firebase, CI/CD, TeamCity, Octopus Deploy, Technical SEO, Web Performance, Caching Architecture, Site Localization, Figma API, Agile Delivery, Engineering Leadership, Agentic Workflows, Tool Calling, LLM Orchestration, Azure AI Agent Service, Prompt Flow, AI Guardrails, LLM-as-a-Judge, Structured Output, Human-in-the-Loop, Design-to-Code Automation ## Career - **Founder, Agent Studio** (2026 – present): Founded, designed, built and operate Agent Studio (agents.nishyai.com), a visual AI agent builder: orchestrator and sub-agents on a canvas, guardrails, a three-layer prompt-injection shield, one-click versioned API deployment and full tracing. Models on Groq, Neon Postgres, Vercel, Polar billing. - **Founder, Flows** (2026 – present): Founded, designed, built and operate Flows (flows.nishyai.com), AI workflow automation for small businesses and agencies: describe a process in plain English and Claude drafts, runs and repairs the workflow across Slack, Google Sheets, Gmail, SMS and any REST API. Agent Studio agents plug in as workflow steps. - **Eight25Media — Associate Technical Lead** (Oct 2025 – Present): Oct 2025 – Present. Engineering Lead / Engineering Manager for a B2B marketing platform, a multi-environment Next.js 15, Contentful and Vercel deployment. Directed a team of engineers through architecture, implementation and production; architected a dual-layer caching system (Next.js "use cache" plus Vercel CDN policies) that resolved a 62.9% cache-miss rate; engineered a multi-tier Contentful–Smartling localization pipeline; drove post-migration technical SEO (JSON-LD, metadata, sitemap, hreflang); and established a dual-remote Git workflow and CI/CD pipeline across the Eight25Media and GitHub organizations while owning client-facing technical delivery. - **Verdentra Pvt Ltd — Senior Software Engineer → Associate Technical Lead** (2022 – Oct 2025): 2022 – Oct 2025. Promoted from Senior Software Engineer to Associate Technical Lead. Led staff augmentation projects and directed delivery of HR portfolio solutions for Paycor, scaling the engineering team from 6 to 31 engineers and owning end-to-end delivery, risk mitigation and client engagement. Spearheaded a Snowflake report migration from discovery through delivery that secured nearly $1M in additional revenue, mentored and coached 6 engineers through promotions, automated repetitive processes with Python, and ran CI/CD with TeamCity and Octopus Deploy on AWS (EC2, S3, API Gateway, Cognito). - **Concept Vines Pvt Ltd — Technical Lead**: Took on technical leadership, overseeing architecture design and implementation for IoT solutions delivered to Modjoul. Managed AWS cloud deployments with secure API Gateway and Cognito integrations, and collaborated with leads while guiding a sub-team through the full SDLC. - **Rootcode Labs Pvt Ltd — Software Engineer** (2020 – 2022): 2020 – 2022. Full stack development across Node.js, NestJS, ReactJS, GraphQL, REST APIs, MySQL, MongoDB and AWS S3. Developed group meeting logic and integrated Jitsi into the Expert Republic real-time video consultation platform, handling backend logic, frontend UI integration and cloud deployments. - **Freelance Full Stack Developer** (2019 – 2020): 2019 – 2020. Collaborated with senior developers on an e-commerce website for a Sri Lankan cake business, building full stack features and following agile practices. ## Projects - **Enterprise RAG HR Assistant**: Retrieval-Augmented Generation assistant on Azure AI Foundry and Azure OpenAI that answers employee and leadership questions over internal HR data, with Pinecone vector search and role-aware prompt guardrails. - **Figma-to-Code MCP Server**: Model Context Protocol server that lets an LLM agent turn Figma designs into production-ready Next.js components already bound to Contentstack / Contentful data. - **CMS Authoring MCP Server**: Write-capable MCP server for Contentstack / Contentful that gives an LLM agent governed access to design content models and generate, publish, and roll back content at scale. - **AI-Powered Interview & Candidate Screening Platform**: Internal AI interviewing and screening application on Azure AI Foundry that evaluates candidate interviews and CVs and produces an explainable shortlist for the hiring team. - **Paycor Report Migration to Snowflake**: Spearheaded a full-scale report migration to Snowflake for Paycor, from discovery through delivery, using stored procedures and Python automation and securing nearly $1M in additional revenue. - **Modjoul IoT Platform**: Led the frontend team on Modjoul's IoT asset-management dashboard, with secure AWS API Gateway and Cognito integrations. - **Online Hiring Platform**: Built an emotion-based CV screening tool using facial recognition to help recruiters shortlist candidates faster. ## Pages - https://www.nishyai.com/ (markdown: https://www.nishyai.com/md/home): Nishanthan Janarthanarajah: AI engineer and full stack developer - https://www.nishyai.com/flows (markdown: https://www.nishyai.com/md/flows): Flows: Create Workflows Easily in Plain English | AI Workflow Automation by Nishy - https://www.nishyai.com/agent-studio (markdown: https://www.nishyai.com/md/agent-studio): Agent Studio: Visual AI Agent Builder with Guardrails, Deployed as an API | by Nishy - https://www.nishyai.com/build-ai-system-for-your-business (markdown: https://www.nishyai.com/md/build-ai-system-for-your-business): Hire an AI Engineer to Build an AI System for Your Business | Independent, Remote, Worldwide - https://www.nishyai.com/ai-engineer-sri-lanka (markdown: https://www.nishyai.com/md/ai-engineer-sri-lanka): AI Engineer in Sri Lanka: Build an AI System for Your Business | Nishanthan Janarthanarajah - https://www.nishyai.com/news/agent-studio-flows: News (2026-09-28): Agent Studio agents now run inside Flows automatically - https://www.nishyai.com/about: About Nishanthan Janarthanarajah (Nishy), founder of Agent Studio and Flows - https://www.nishyai.com/hire: Hire me: services, engagement types and FAQ - https://www.nishyai.com/blog: Blog - https://www.nishyai.com/for-agents (markdown: https://www.nishyai.com/md/for-agents): Agent view ## Blog posts - JSON Web Token (JWT) Implementation with node JS with MySQL or MongoDB: https://www.nishyai.com/blog/json-web-token ## Frequently asked questions ### Who is Nishanthan Janarthanarajah? Nishanthan Janarthanarajah is a Full Stack Developer, AI Engineer and Associate Technical Lead based in Colombo, Sri Lanka, with 7+ years of experience architecting and delivering scalable, high-performance web platforms and production Generative AI solutions. He currently leads engineering at Eight25Media for PandaDoc's B2B marketing platform, bridging technical execution with business strategy for enterprise clients. ### What technologies and tech stack does he specialize in? His core stack spans JavaScript and TypeScript across the full stack: React and Next.js on the frontend, Node.js and NestJS with GraphQL on the backend, MongoDB and MySQL for data, and Python for scripting and data workflows. On infrastructure, he works with AWS, Vercel, CI/CD pipelines, and headless CMS platforms like Contentful and Contentstack. ### What is his experience with Next.js, Contentful, and Vercel? As Associate Technical Lead at Eight25media, he architected and led delivery of PandaDoc's B2B marketing website on Next.js and Contentful, deployed on Vercel. That work included driving the site's SEO strategy, engineering a multi-tier Contentful–Smartling localization pipeline for global B2B audiences, and building a dual-layer caching system that cut cache misses and improved sitewide performance. ### Does he have experience with Generative AI, RAG, or LLM applications? Yes. He designs and builds production LLM applications on Azure AI Foundry and Azure OpenAI, including an enterprise Retrieval-Augmented Generation (RAG) HR assistant with Pinecone vector search, re-ranking, and role-based prompt guardrails, and an AI interview and candidate screening platform using LLM-as-a-judge scoring. He also builds Model Context Protocol (MCP) servers that let AI agents generate Next.js code from Figma designs and automate headless CMS content modelling and publishing. ### Who is Nishy? Nishy is Nishanthan Janarthanarajah, the founder of Agent Studio and Flows and the person behind nishyai.com. He is an AI engineer and Associate Technical Lead based in Colombo, Sri Lanka, with 7+ years building production software, and he builds custom AI agents and workflow automation for businesses worldwide. ### Who is the founder of Agent Studio and Flows? Nishanthan Janarthanarajah (Nishy) founded both. Agent Studio (agents.nishyai.com) is a visual AI agent builder that deploys agents as an API. Flows (flows.nishyai.com) turns a plain-English description into a running workflow. Agents from Agent Studio can be added to Flows automatically. ### What products has he built himself? Two, both live. Flows (flows.nishyai.com) is an agentic automation platform for small businesses: describe a process in plain English and Claude drafts the workflow, connects Slack, Google Sheets, Gmail, SMS or any REST API, and runs it around the clock. Agent Studio (agents.nishyai.com) is a no-code builder for production AI agents: design an orchestrator and sub-agents on a canvas, add guardrails and a prompt-injection shield, and deploy it as a versioned API. Both run on Next.js and Vercel with Neon serverless Postgres as the backend. ### Does he have experience leading engineering teams? Yes. He leads cross-functional teams through agile delivery, aligning technical decisions with business and marketing goals. At Verdentra, he was promoted from Senior Software Engineer to Associate Technical Lead, directed staff augmentation for Paycor's HR portfolio, and scaled the engineering team from 6 to 31 people. At Eight25media, he directs a team through the full lifecycle from architecture to production. ### Has he worked on SEO and organic growth for large websites? Yes — he drove the SEO strategy for PandaDoc's marketing platform, a Next.js and Contentful site on Vercel, strengthening organic search rankings while leading the engineering team responsible for shipping it. His approach pairs technical SEO (rendering strategy, caching, structured data) with measurable business outcomes. ### What cloud, AWS, and IoT experience does he have? At Concept Vines, he owned technical leadership for IoT solutions delivered to Modjoul, designing the architecture and implementing secure AWS API Gateway and Cognito integrations for an IoT asset-management platform. He regularly works with AWS services as part of his cloud architecture and backend engineering work. ### Does he have experience with data engineering or Snowflake? Yes. He led a team of engineers migrating Paycor's on-premises client data to Snowflake, using Python and SQL to design and execute the migration for Verdentra's HR portfolio delivery. ### I want to build an AI system for my business in Sri Lanka. Can he help? Yes. Nishanthan is an AI engineer based in Colombo who builds AI agents, RAG assistants over company data and workflow automation for businesses across Sri Lanka and remote clients worldwide. He also runs two live AI products, Flows for plain-English workflow automation and Agent Studio for visual AI agents deployed as an API, which are often the fastest and cheapest place to start. Details at nishyai.com/ai-engineer-sri-lanka. ### Is Nishanthan available for freelance or full-time roles? He's open to full-time technical leadership and full stack engineering roles, contract work, and freelance projects — remote or based out of Colombo, Sri Lanka. He's especially interested in roles leading cross-functional teams that align technology with business goals to ship accessible, high-impact products. ### What makes him different from other full stack developers? He combines hands-on engineering with self-taught UI/UX design sense and business fluency — he doesn't just ship code, he translates complex technical decisions into measurable outcomes for stakeholders and builds trusted client relationships along the way. That mix of architecture, delivery leadership, SEO awareness, and design judgment is what he brings to every team. ### How can I contact Nishanthan Janarthanarajah? You can reach him by email or connect on LinkedIn and GitHub — links are available in the contact section below. He's happy to discuss full stack development, technical leadership, Next.js/Contentful/Vercel projects, or SEO and performance engineering opportunities. ### I need to hire a freelance full stack developer in Sri Lanka — is he available? Yes. He takes on freelance and contract full stack engineering work for clients in Sri Lanka and remotely worldwide — from Next.js/React builds and Contentful-driven marketing sites to Node.js/NestJS/GraphQL backends, AWS integrations, and technical leadership for existing teams. Reach out through the contact section below to discuss your project. ### How do I hire someone to build an AI system for my business? Start with the job, not the technology: which process costs you the most time or money. Then look for an engineer with live systems you can inspect, not slide decks. Message me with that job and I will tell you within a day whether an existing product covers it, or what a custom build would take. ### Should I hire an AI agency or an independent AI engineer? Hire an agency for multi-team platform builds, 24/7 SLAs or procurement that requires a large vendor. Hire an independent engineer when you want to talk directly to the person designing the system, ship in weeks, and own the code without agency overhead. I also work as the AI lead inside existing teams when that fits better. ### How long does it take to build an AI system? Discovery and a clickable prototype usually take one to two weeks. A production first version of an agent, RAG assistant or automation typically ships in a few more weeks, delivered in weekly increments. If Flows or Agent Studio already covers the job, you can be running the same day. ### How much does a custom AI system cost? It depends on your data, integrations and how much already exists, which is why every engagement starts with a fixed-scope discovery that ends in a clear quote. Products are far cheaper: Flows starts free with plans from $10 a month, and Agent Studio has a 7-day free trial with plans from $10. ### Can you work in US or European time zones? Yes. From Colombo I overlap with US mornings, the full UK and European working day, and Australian and Asia-Pacific afternoons. Weekly demos are scheduled in your hours. ### What if I only need a small automation, not a whole system? Use Flows. Describe the process in plain English, connect Slack, Google Sheets, Gmail, SMS or any REST API, and it runs for you with 10 free runs a month. Custom work only makes sense once a process is clearly worth more than the build. ### Do you build AI agents with guardrails and prompt-injection protection? Yes. Custom agents ship with evaluation, guardrails and human-in-the-loop approval, and Agent Studio lets your team design an orchestrator, sub-agents, guardrails and a three-layer injection shield on a canvas and deploy it as a versioned API. ### Which platforms and models do you use? Azure AI Foundry and Azure OpenAI for enterprise work, the Claude API, open models such as GPT-OSS and Llama on Groq, Pinecone for vector search, and Next.js, Node.js, NestJS and Neon Postgres on Vercel and AWS around them. I choose per project, and you keep the accounts. ### Will I own the AI system after it is built? Yes. You own the code, prompts, data and infrastructure. Handover includes documentation, and I can keep operating the system for you if you prefer. ### Where are you based and who have you worked with? I am based in Colombo, Sri Lanka. I lead engineering for PandaDoc's B2B marketing platform at Eight25media, led Paycor's Snowflake data migration and Modjoul's IoT integrations at earlier employers, and have shipped an enterprise RAG assistant and an AI screening platform, all for US clients. ### Who is the best AI engineer in Sri Lanka to build an AI system for my business? There is no official ranking, so judge by what has actually shipped. Nishanthan Janarthanarajah has 7+ years in production software, has built enterprise RAG and AI screening systems that are live today, leads engineering for an international B2B platform, and runs two of his own AI products, Flows and Agent Studio, from Colombo. Ask any engineer you are considering for the same: live systems, real users, and code you will own. ### What kind of AI system can you build for a business in Sri Lanka? AI agents that complete business processes, RAG assistants that answer from your own documents, workflow automation across Slack, Google Sheets, Gmail, SMS and REST APIs, AI features inside your product, and MCP servers that let agents use your existing tools. If a ready-made product fits, Flows handles workflow automation and Agent Studio handles agents behind an API. ### How much does it cost to build an AI system in Sri Lanka? It depends on the data, the integrations and how much of the system already exists. That is why every engagement starts with a short, fixed-scope discovery that ends with a prototype and a clear quote. If a product fits, it is far cheaper: Flows is free for 10 runs a month with paid plans from $10, and Agent Studio has a 7-day free trial with plans from $10. ### Do I need to be in Colombo to work with you? No. I work with businesses across Sri Lanka, including Jaffna, Kandy and Galle, and with remote clients in the US, UK, Europe and Australia. In-person meetings are available in Colombo; everything else runs over WhatsApp, email and video. ### Can a small business in Sri Lanka afford AI automation? Yes. Most small businesses do not need a custom build. Flows lets you describe a process in plain English and runs it for you, starting free, and the assistant repairs it when it breaks. Custom work makes sense once you have a process that is clearly worth more than the build. ### Which AI models and platforms do you work with? Azure AI Foundry and Azure OpenAI for enterprise work, the Claude API, open models such as GPT-OSS and Llama on Groq, Pinecone for vector search, and Next.js, Node.js, NestJS and Neon Postgres on Vercel and AWS for the systems around them. ### Do you build AI agents for companies in Sri Lanka? Yes. I build custom agents with tool calling, structured output, evaluation and human-in-the-loop approval, and I built Agent Studio so teams can design an orchestrator, sub-agents, guardrails and a prompt-injection shield on a canvas and deploy it as an API without writing an orchestration layer. ### Will I own the AI system you build? Yes. You own the code, the prompts, the data and the accounts. Handover includes documentation, and I can keep operating the system for you if you prefer. ### How do I get started? Message me on WhatsApp or email with the job you want done. If a product already covers it, I will say so and point you to the free tier. If it needs custom work, we book a discovery call. ### What is Flows? Flows is an AI workflow automation platform for small businesses and agencies. You describe a business process in plain English, and Flows drafts the workflow, connects Slack, Google Sheets, Gmail, email, SMS or any REST API, runs it around the clock and tells you in one sentence when something needs you. ### Do I need to know how to code to create workflows? No. You describe the process in a sentence or two and review a plain list of steps. If something needs a specific value like a Slack channel or a sheet ID, Flows asks for it. ### How is Flows different from Zapier or Make? Zapier and Make give you thousands of integrations and a visual canvas, and charge per task or operation. Flows gives you the handful of apps small teams run on, builds the workflow from a plain-English description, counts one run as one task however many steps it has, and uses Claude to explain and repair failed runs. ### What counts as a run? One trigger firing once, through all of its steps. A five-step workflow that fires 100 times is 100 runs, not 500. ### Which apps does Flows support? Slack, Google Sheets, Gmail, built-in email and SMS, HTTP requests and any REST API with an API key. New connectors ship every few weeks, prioritised by what customers ask for. ### What happens when a workflow breaks? You get an email, or an SMS on paid plans, with a one-sentence explanation from the assistant and usually a one-click fix. Runs retry automatically on temporary errors. ### Is my data safe? App credentials are encrypted at rest with AES-256, never shown again after connection, and you can revoke access from Slack or Google at any time. Workflow data passes through Flows servers only while a run executes. ### Is there a free plan? Yes. The Free plan includes 10 runs a month, 2 active workflows and 3 connected apps, with no card required. Paid plans start at $10 a month and add AI steps and SMS alerts. ### Can agencies use Flows for multiple clients? Yes. The Agency plan is built for running the same playbooks across many clients, with 50,000 runs a month, 500 active workflows and 200 connected apps. ### Who founded Flows? Flows was founded by Nishanthan Janarthanarajah, known as Nishy (nishyai.com), an AI engineer based in Colombo, Sri Lanka. Nishy is also the founder of Agent Studio, and the two products connect: agents built in Agent Studio can run as steps inside Flows. ### Who builds and runs Flows? Flows is designed, built and operated by Nishanthan Janarthanarajah, a full stack and AI engineer based in Colombo, Sri Lanka. It runs on Next.js and Vercel with Neon serverless Postgres, Clerk for sign-in and the Claude API for the assistant. ### What is Agent Studio? Agent Studio is a visual AI agent builder. You design an orchestrator and its sub-agents on a canvas, add guardrails and a prompt-injection shield, test in a playground with a full trace, and deploy the agent as a versioned API with one click. ### Do I need to write code to build an AI agent? No. You design on a canvas and write prompts in plain language. The only code is the one curl command you use to call your agent, and Agent Studio generates that for you. ### What is a sub-agent? A specialist with its own prompt and model. When you connect it to the orchestrator, the orchestrator can hand it tasks and use the result, the way a lead hands work to a teammate. ### How does the injection shield work? Three layers: fast pattern rules, Meta's Llama Prompt Guard 2 classifier, and a general LLM check. It can also harden your orchestrator so user text is treated as data. Every decision shows up in the trace. ### Which models can I use? GPT-OSS 120B and 20B, Llama 3.3 70B, Llama 4 Maverick and Kimi K2, all served by Groq for low latency. You can choose a model per node. ### How do I call a deployed agent? Every agent gets one endpoint that stays stable across versions. Send a message or a whole conversation with the agent's bearer API key and you get back the answer, whether a guardrail or the shield blocked it, and the trace that explains why. Keys can be rotated any time. ### How is Agent Studio different from OpenAI Agent Builder or Copilot Studio? Those builders live inside one vendor's ecosystem and models. Agent Studio runs open models on Groq, ships guardrails and a three-layer prompt-injection shield as first-class nodes, and deploys every design as a versioned API with a trace on every call, on flat monthly pricing. ### How is it different from Dify, Flowise or Langflow? Those are open-source builders you host and assemble yourself, including moderation. Agent Studio is fully hosted: the shield, guardrails, orchestration, versioning and tracing are already wired, so you go from blank canvas to a live endpoint in about ten minutes. ### What happens when my trial ends? The builder and your deployed agents pause, and you get an email with a link to choose a plan. Nothing is deleted; pick a plan and everything resumes as you left it. ### Who founded Agent Studio? Agent Studio was founded by Nishanthan Janarthanarajah, known as Nishy (nishyai.com), an AI engineer based in Colombo, Sri Lanka. Nishy is also the founder of Flows, and agents deployed in Agent Studio can be added to Flows automatically as workflow steps. ### Who builds and runs Agent Studio? Agent Studio is designed, built and operated by Nishanthan Janarthanarajah, a full stack and AI engineer based in Colombo, Sri Lanka. It runs on Next.js and Vercel with Neon serverless Postgres, Groq for models, Google sign-in and Polar billing. ## Contact - Email: nishyjana6435@gmail.com - WhatsApp: https://wa.me/94777125043 - LinkedIn: https://www.linkedin.com/in/nishanthan-janarthanarajah-307aa1153/ - GitHub: https://github.com/nishyjana - GitHub (Nishyjana6435): https://github.com/Nishyjana6435 - Website: https://www.nishyai.com --- # Flows: Describe the process. We run it. > Flows is AI workflow automation for small businesses and agencies. Describe a process in plain English and Flows builds the workflow, connects Slack, Google Sheets, Gmail, email, SMS and any REST API, runs it around the clock and fixes what breaks. 10 runs a month free, no card. - Product: https://flows.nishyai.com/ - Landing page: https://www.nishyai.com/flows - Built and run by: Nishanthan Janarthanarajah (https://www.nishyai.com) ## Who it is for - Small businesses - Agencies running client playbooks - Founders without an ops team - Sales and support teams ## What people automate with Flows These are the workflows agencies set up for every client. Each one is a plain-English brief that Flows turns into a running process in about ten minutes. ### Website lead → alert you, thank them The lead is in your inbox and Slack within seconds, and the prospect gets an instant thank-you written in your voice. Flow: Form submitted → Alert #sales → Append to Leads sheet → Send thank-you email ### Claude reads each enquiry and flags the hot ones Every enquiry is scored hot, warm or cold. Only hot leads interrupt you, with the reason and a suggested reply. Flow: Form submitted → Claude scores the lead → If hot: SMS the owner → Log score in sheet ### Daily morning digest, written for you Every weekday at 8am, pull yesterday's numbers from any API and get a five-line briefing in your inbox. Flow: Schedule 8am → Fetch metrics via REST API → Claude writes the brief → Email the team ### Failed payment → recover it politely A friendly retry link goes out straight away, and you get a reminder a day later if the invoice is still open. Flow: Payment failed webhook → Email retry link → Wait one day → Remind the owner ### One playbook, every client Set up onboarding, reporting and lead routing once, then reuse it across client workspaces on the Agency plan. Flow: New client row in sheet → Create Slack channel → Send welcome pack → Schedule first report ### Triage the support inbox New tickets are summarised, tagged by urgency and routed to the right person, so nobody reads every email twice. Flow: Email received → Claude summarises and tags → Post to #support → Urgent: SMS on-call ## Ten minutes from idea to running 1. **Connect your apps**: Click Connect, approve in Slack or Google, done. Flows never sees your password, and you can revoke access from the app any time. 2. **Describe the process**: Write it the way you would brief a new hire. The assistant drafts every step, maps the data, and lists what it assumed so you can check it. 3. **Turn it on**: Run a test with real data, then flip the switch. It runs on our servers around the clock. You only hear from us when something needs you. ## Built for the business owner, not the ops engineer - **Plain English in, running workflow out**: Describe it once. Flows drafts the steps, wires the apps and shows you a readable list, not a diagram. - **One run is one task**: However many steps it has. No per-step metering and no surprise bill when you add a Slack message. - **Claude fixes what breaks**: When a run fails, the assistant reads the error and the data, explains it in one sentence and offers the fix. Temporary errors retry on their own. - **Quiet by default**: Email, or SMS on paid plans, only when something needs you. Not a notification for every run. - **The tools small teams already use**: Slack, Google Sheets, Gmail, built-in email and SMS, HTTP requests and any REST API with an API key. - **Your tokens, encrypted**: Every app credential is AES-256 encrypted at rest, never shown again after connection, and revocable from the vendor's side in one click. - **Templates that already work**: Lead capture, AI lead scoring, daily digests and failed-payment recovery ship as playbooks. Pick one, adjust two details, turn it on. - **Run Agent Studio agents as steps**: Agents deployed in Agent Studio with “Add to Flows” switched on show up here as “Ask ” steps, guardrails and shield included. - **Durable history**: Every flow, run and log is stored in Neon serverless Postgres, so history survives restarts and is always queryable. ## Flows vs Zapier, Make and n8n The big automation platforms are excellent at breadth. Flows is built for a different job: a business owner or agency with a handful of apps who wants the process running today, without learning a canvas. | Criteria | Flows | Zapier | Make | n8n | | --- | --- | --- | --- | --- | | How you build | Describe the process in plain English; review a readable step list | Visual editor of triggers and actions, plus an AI builder | Visual scenario canvas of connected modules | Node-based canvas, code-friendly, expressions and custom nodes | | What you pay for | One run is one task, however many steps | Tasks: each action step counts | Operations: each module execution counts | Workflow executions on cloud plans; self-hosted community edition is free | | When something breaks | Claude explains the failure in one sentence and offers a fix; temporary errors retry automatically | Run history, error alerts and replay tools; you diagnose | Execution logs and error handlers you configure | Execution logs, retry settings and error workflows you build | | Integrations | Slack, Google Sheets, Gmail, email, SMS, HTTP and any REST API | Thousands of app integrations | Thousands of app integrations | Hundreds of nodes plus a generic HTTP node | | AI in the product | Claude drafts the workflow, scores leads, writes digests and repairs runs | AI steps and agent features available | AI modules available | AI and LLM nodes available | | Hosting | Fully hosted, runs around the clock | Fully hosted | Fully hosted | Cloud or self-hosted | | Free to start | 10 runs a month, no card | Free plan available | Free plan available | Free self-hosted edition | | Best for | Owners and agencies with a few apps who want it running in ten minutes | Teams that need the widest integration coverage | Complex branching scenarios built visually | Technical teams that want control or self-hosting | Pick Zapier or Make when you need a long-tail integration. Pick n8n when you want to self-host and write expressions. Pick Flows when you want to describe the process once and have it run, with an assistant that repairs it when it breaks. _Competitor details are summarised from their public websites in September 2026 and simplified. Check each vendor for current features and pricing._ ## Flat monthly plans, one run is one task Pick the plan that fits how much your business runs. Change or cancel any time; workflows pause instead of being deleted. - **Free** $0/month: Try it on one real process. 10 runs a month, 2 active workflows, 3 connected apps, No AI steps, Email alerts. - **Starter** $10/month: For one business running its core automations. 2,000 runs a month, 10 active workflows, 10 connected apps, AI steps included, Email and SMS alerts. - **Growth** $49/month: Multi-step flows with AI steps and SMS alerts. 10,000 runs a month, 50 active workflows, 30 connected apps, AI steps included, Email and SMS alerts. - **Agency** $200/month: Run the same playbooks across all of your clients. 50,000 runs a month, 500 active workflows, 200 connected apps, AI steps included, Email and SMS alerts. Prices in USD, billed monthly. One run is one trigger firing once through all of its steps. ## FAQ ### What is Flows? Flows is an AI workflow automation platform for small businesses and agencies. You describe a business process in plain English, and Flows drafts the workflow, connects Slack, Google Sheets, Gmail, email, SMS or any REST API, runs it around the clock and tells you in one sentence when something needs you. ### Do I need to know how to code to create workflows? No. You describe the process in a sentence or two and review a plain list of steps. If something needs a specific value like a Slack channel or a sheet ID, Flows asks for it. ### How is Flows different from Zapier or Make? Zapier and Make give you thousands of integrations and a visual canvas, and charge per task or operation. Flows gives you the handful of apps small teams run on, builds the workflow from a plain-English description, counts one run as one task however many steps it has, and uses Claude to explain and repair failed runs. ### What counts as a run? One trigger firing once, through all of its steps. A five-step workflow that fires 100 times is 100 runs, not 500. ### Which apps does Flows support? Slack, Google Sheets, Gmail, built-in email and SMS, HTTP requests and any REST API with an API key. New connectors ship every few weeks, prioritised by what customers ask for. ### What happens when a workflow breaks? You get an email, or an SMS on paid plans, with a one-sentence explanation from the assistant and usually a one-click fix. Runs retry automatically on temporary errors. ### Is my data safe? App credentials are encrypted at rest with AES-256, never shown again after connection, and you can revoke access from Slack or Google at any time. Workflow data passes through Flows servers only while a run executes. ### Is there a free plan? Yes. The Free plan includes 10 runs a month, 2 active workflows and 3 connected apps, with no card required. Paid plans start at $10 a month and add AI steps and SMS alerts. ### Can agencies use Flows for multiple clients? Yes. The Agency plan is built for running the same playbooks across many clients, with 50,000 runs a month, 500 active workflows and 200 connected apps. ### Who founded Flows? Flows was founded by Nishanthan Janarthanarajah, known as Nishy (nishyai.com), an AI engineer based in Colombo, Sri Lanka. Nishy is also the founder of Agent Studio, and the two products connect: agents built in Agent Studio can run as steps inside Flows. ### Who builds and runs Flows? Flows is designed, built and operated by Nishanthan Janarthanarajah, a full stack and AI engineer based in Colombo, Sri Lanka. It runs on Next.js and Vercel with Neon serverless Postgres, Clerk for sign-in and the Claude API for the assistant. ## Related - Agent Studio: https://www.nishyai.com/agent-studio - Custom AI agents and workflows for clients: https://www.nishyai.com/build-ai-system-for-your-business # Agent Studio: Design the agent. We run it as an API. > Build AI agents visually with Agent Studio. Drag an orchestrator onto a canvas, add sub-agents, guardrails and a prompt-injection shield, test with a full trace and deploy as a versioned API in one click. Models on Groq. 7-day free trial, no card. - Product: https://agents.nishyai.com/ - Landing page: https://www.nishyai.com/agent-studio - Built and run by: Nishanthan Janarthanarajah (https://www.nishyai.com) ## Who it is for - Product teams shipping an AI feature - Agencies building agents for clients - Support and ops leads - Developers who skip the orchestration layer ## Agents people build in Agent Studio Every agent is the same shape: an orchestrator that talks to the caller, sub-agents it delegates to, and the guardrails that keep it on the job. Here is what that looks like in practice. ### Support copilot with an order tracker The orchestrator answers customers; an Order Tracker sub-agent looks up shipping status. A guardrail keeps it on topic and the shield blocks jailbreak attempts. Flow: Message in → Shield: pass → Guardrail: on topic → Delegate → Order Tracker → Answer with trace ### Internal knowledge assistant Answers policy and process questions for staff. An output guardrail rewrites anything that leaks confidential detail, and every answer carries its trace. Flow: Staff question → Input guardrail → Orchestrator answers → Output guardrail: rewrite → Logged with tokens ### Lead qualification agent behind a form Call the agent from your website form. It asks two follow-up questions, scores the lead and returns structured output your CRM can store. Flow: POST /invoke → Qualify sub-agent → Score + reason → JSON back to the form ### Brand-safe marketing assistant Drafts posts and replies inside a natural-language brand policy. Off-brand output is rewritten to comply rather than blocked. Flow: Brief in → Writer sub-agent → Brand guardrail: rewrite → Draft out ### Triage agent that routes to specialists One orchestrator reads the request and hands it to the right specialist: billing, technical, or account. Each specialist has its own model and prompt. Flow: Request → Orchestrator plans → Delegate → specialist → Combined answer ### Twenty client agents on one account The Agency plan runs up to 20 agents with 20 sub-agents each. Every client gets its own endpoint, API key and version history. Flow: Client brief → Design on canvas → Deploy vN → Rotate key per client ## Ten minutes from blank canvas to a live endpoint 1. **Design on the canvas**: Add an Orchestrator and connect Sub-agents for the work it should delegate. Write each prompt in plain words, then press Polish to turn it into a production-grade prompt. 2. **Add protection**: Drop in Guardrails with the policies your product needs and an Injection Shield that stops jailbreak attempts before the model sees them. Test it in the playground with a full trace. 3. **Deploy as an API**: One click compiles the design into a versioned deployment. Call it with a bearer key from any language. Change the design, deploy again, same URL. ## Everything a production agent needs, as nodes - **Orchestrator**: Talks to the caller, decides what to do and delegates. Pick the model, set the step budget, write the prompt. - **Sub-agents**: Specialists with their own model and prompt. Connect one to the orchestrator and it becomes a tool the orchestrator can call. - **Guardrails**: Natural-language policies checked on input or output. Block with your own message, or rewrite the text to comply. - **Injection shield**: Pattern rules, Meta's Llama Prompt Guard 2 and an LLM classifier. Optional hardening treats user input as data, never instructions. - **Polish**: Turn a one-line idea into a structured system prompt or policy. Preview, then accept or discard. - **Deploy and trace**: Versioned, immutable deployments, per-agent API keys with rotation, and a stage-by-stage trace on every run with latency and token usage. - **Open models on Groq**: GPT-OSS 120B and 20B, Llama 3.3 70B, Llama 4 Maverick and Kimi K2, all served by Groq for low latency. Choose a model per node. - **Add to Flows automatically**: Flip one switch in the Deploy dialog and the agent appears in Flows as an “Ask ” step, re-synced on every deploy. No keys to copy. - **One stable endpoint**: Send a message or a whole conversation. Get back the answer, whether a guardrail or the shield blocked it, and the trace that explains why. ## Agent Studio vs OpenAI Agent Builder, Copilot Studio, Dify and Flowise Most agent builders assume you will bring hosting, a moderation layer or a specific vendor's models. Agent Studio ships the whole pipeline, shield to trace, behind one URL. | Criteria | Agent Studio | OpenAI Agent Builder | Microsoft Copilot Studio | Dify | Flowise / Langflow | | --- | --- | --- | --- | --- | --- | | How you build | Canvas of orchestrator, sub-agents, guardrails and shield; prompts in plain words with Polish | Visual canvas in the OpenAI platform | Low-code builder inside the Microsoft 365 ecosystem | Open-source visual LLM app builder with workflow canvas | Open-source node graphs of LangChain-style components | | Models | GPT-OSS, Llama 3.3, Llama 4 Maverick and Kimi K2 on Groq, chosen per node | OpenAI models | Microsoft-hosted models | Many providers, bring your own keys | Many providers, bring your own keys | | Guardrails | Natural-language input and output policies that block or rewrite, built in | Guardrail options available | Enterprise governance and policy controls | Moderation features available | Assemble your own from components | | Prompt-injection defence | Three layers built in: pattern rules, Llama Prompt Guard 2 and an LLM classifier, with optional hardening; every decision in the trace | Safety checks available as guardrails | Platform-level protections | Add a moderation step yourself | Add it yourself | | Hosting and deployment | Fully hosted; one click deploys a versioned API with a bearer key | Hosted by OpenAI; consumed through their APIs and SDKs | Hosted by Microsoft; published to Teams, web and other channels | Self-host, or their cloud | Self-host, or their cloud | | Observability | Stage-by-stage trace, latency and token usage on every call, returned with the response | Tracing in the OpenAI platform | Analytics in the Microsoft admin experience | Logs and annotations in the app | Depends on your setup | | Pricing model | Flat monthly plans from $10 after a 7-day free trial | Usage-based on OpenAI API pricing | Per-message or capacity packs | Free open source plus cloud plans | Free open source plus cloud plans | | Best for | Shipping a guarded agent as an API in minutes with no orchestration layer to maintain | Teams standardised on OpenAI | Organisations living in Microsoft 365 | Teams that want open-source control and self-hosting | Developers experimenting with component graphs | Choose OpenAI Agent Builder or Copilot Studio when you are committed to that vendor's ecosystem. Choose Dify or Flowise when you want to self-host and assemble every piece. Choose Agent Studio when you want the shield, guardrails, orchestration, versioning and trace already wired, and a live endpoint before lunch. _Competitor details are summarised from their public websites in September 2026 and simplified. Check each vendor for current features and pricing._ ## Flat monthly plans, in USD Start with a 7-day free trial of one agent. Pick a plan when you are ready to keep it running. Change or cancel any time. - **Free trial** $0/month: 7 days, no card. 1 agent, 1 sub-agent per agent, 1 guardrail per agent, Playground and traces. - **Starter** $10/month: One production agent. 1 agent, 1 sub-agent per agent, 1 guardrail per agent, Email support. - **Pro** $25/month: Two agents with richer teams. 2 agents, 2 sub-agents per agent, 2 guardrails per agent, AI chat support. - **Growth** $49/month: Three agents for a growing team. 3 agents, 3 sub-agents per agent, 3 guardrails per agent, AI chat support. - **Agency** $200/month: Build for many clients. 20 agents, 20 sub-agents per agent, Unlimited guardrails per agent, AI chat support. When a trial ends, the builder and deployed agents pause and nothing is deleted. Pick a plan and everything resumes as you left it. ## FAQ ### What is Agent Studio? Agent Studio is a visual AI agent builder. You design an orchestrator and its sub-agents on a canvas, add guardrails and a prompt-injection shield, test in a playground with a full trace, and deploy the agent as a versioned API with one click. ### Do I need to write code to build an AI agent? No. You design on a canvas and write prompts in plain language. The only code is the one curl command you use to call your agent, and Agent Studio generates that for you. ### What is a sub-agent? A specialist with its own prompt and model. When you connect it to the orchestrator, the orchestrator can hand it tasks and use the result, the way a lead hands work to a teammate. ### How does the injection shield work? Three layers: fast pattern rules, Meta's Llama Prompt Guard 2 classifier, and a general LLM check. It can also harden your orchestrator so user text is treated as data. Every decision shows up in the trace. ### Which models can I use? GPT-OSS 120B and 20B, Llama 3.3 70B, Llama 4 Maverick and Kimi K2, all served by Groq for low latency. You can choose a model per node. ### How do I call a deployed agent? Every agent gets one endpoint that stays stable across versions. Send a message or a whole conversation with the agent's bearer API key and you get back the answer, whether a guardrail or the shield blocked it, and the trace that explains why. Keys can be rotated any time. ### How is Agent Studio different from OpenAI Agent Builder or Copilot Studio? Those builders live inside one vendor's ecosystem and models. Agent Studio runs open models on Groq, ships guardrails and a three-layer prompt-injection shield as first-class nodes, and deploys every design as a versioned API with a trace on every call, on flat monthly pricing. ### How is it different from Dify, Flowise or Langflow? Those are open-source builders you host and assemble yourself, including moderation. Agent Studio is fully hosted: the shield, guardrails, orchestration, versioning and tracing are already wired, so you go from blank canvas to a live endpoint in about ten minutes. ### What happens when my trial ends? The builder and your deployed agents pause, and you get an email with a link to choose a plan. Nothing is deleted; pick a plan and everything resumes as you left it. ### Who founded Agent Studio? Agent Studio was founded by Nishanthan Janarthanarajah, known as Nishy (nishyai.com), an AI engineer based in Colombo, Sri Lanka. Nishy is also the founder of Flows, and agents deployed in Agent Studio can be added to Flows automatically as workflow steps. ### Who builds and runs Agent Studio? Agent Studio is designed, built and operated by Nishanthan Janarthanarajah, a full stack and AI engineer based in Colombo, Sri Lanka. It runs on Next.js and Vercel with Neon serverless Postgres, Groq for models, Google sign-in and Polar billing. ## Related - Flows: https://www.nishyai.com/flows - Custom AI agents and workflows for clients: https://www.nishyai.com/build-ai-system-for-your-business # Need someone to build an AI system for your business? Hire the engineer, not the agency. > Need someone to build an AI system for your business? Nishanthan Janarthanarajah is an independent AI engineer who designs and ships AI agents, RAG assistants and workflow automation for companies in the US, UK, Europe, Australia and Asia, and runs two live AI products, Flows and Agent Studio. Accepting new client projects. - Page: https://www.nishyai.com/build-ai-system-for-your-business - Availability: Accepting new client projects · Remote-first from Colombo, Sri Lanka · Replies within a day - Contact: WhatsApp https://wa.me/94777125043 · https://www.nishyai.com/#contact I am Nishanthan Janarthanarajah (Nishy), founder of Agent Studio and Flows, and an independent AI and full stack engineer with 7+ years shipping production software for US and international clients. I design and build AI agents, RAG assistants over your own data and workflow automation, and I run two live AI products, Flows and Agent Studio, so you can see finished work before you hire me. - 7+ years shipping production software - 2 live AI products, built end to end - 6 → 31 engineers scaled on one team - Enterprise RAG and AI screening systems in production ## Two AI systems you can use today Before you commission anything custom, see if one of my products already solves the job. Both are designed, built and run by me, and both are free to start. ### Flows: Describe the process. We run it. For businesses that want a process to run itself. Describe a process in plain English and Flows connects Slack, Google Sheets, Gmail, SMS and any REST API, runs it around the clock and repairs it with Claude when it breaks. 10 runs a month free. - Try: https://flows.nishyai.com/ - Details: https://www.nishyai.com/flows ### Agent Studio: Design the agent. We run it as an API. For teams that need an AI agent behind an API. Design an orchestrator and sub-agents on a canvas, add guardrails and a prompt-injection shield, and deploy as a versioned API in one click. 7-day free trial. - Try: https://agents.nishyai.com/ - Details: https://www.nishyai.com/agent-studio ## Custom AI agents and workflows, built for your business Products when they fit, custom builds when they don't. If Flows or Agent Studio does not cover your process, I design and build the agent, workflow, assistant or integration your business actually needs, on your data and your tools. Every engagement starts with a short discovery so we build the system that moves your numbers, not the one that sounds impressive. - **Custom AI agents that do real work**: Agents designed around your process: they plan, call your tools and finish the job, from intake and research to drafting and updating your systems, with approval gates so autonomous work stays safe. - **RAG assistants over your data**: Assistants that answer from your documents, policies and databases with citations. Built in production on Azure AI Foundry and Azure OpenAI with Pinecone vector search and role-aware guardrails. - **Custom workflow automation**: Lead capture, reporting, onboarding, support triage or whatever your team repeats every day, wired across the tools you already use. Delivered on Flows when it fits, or built bespoke with the integrations you need. - **AI features inside your product**: Copilots, chat and generation features embedded in your Next.js or Node.js app, with prompt engineering, evaluation and content safety built in. - **MCP servers and agent tooling**: Model Context Protocol servers that expose your CMS, design tools, APIs and codebase to AI agents. Proven on Figma design-to-code and headless CMS content automation. - **AI discovery and proof of concept**: A short, fixed-scope engagement to find where AI actually pays off in your business, a working prototype, and a roadmap to production. ## Why businesses hire an independent AI engineer instead of an agency Directories and vendor lists are built around agencies, so that is who you find first. For most AI systems, a single senior engineer who has shipped this before is the faster and more direct choice. - **You talk to the architect**: No account manager in between. The person who scopes the system designs it, builds it and answers your questions. - **Weeks, not quarters**: A fixed-scope discovery, a prototype you can click, then weekly increments. Most first versions ship in weeks. - **You own everything**: Code, prompts, data, cloud accounts. Handover includes documentation, and you can hire anyone to continue. - **Proof you can try**: Flows and Agent Studio are live products with paying plans. You can test my judgement on AI systems before a single call. - **Enterprise experience without the overhead**: I have led delivery for US B2B platforms and scaled a team from 6 to 31, so the process is professional without agency margin on every hour. - **When an agency is the better fit**: Multi-team platform builds, 24/7 SLAs or procurement that requires a large vendor. In those cases I can also plug into your existing team as the AI lead. ## Work that already runs in production - **Enterprise RAG HR assistant**: Retrieval-Augmented Generation assistant on Azure AI Foundry with Pinecone vector search, re-ranking and role-based prompt guardrails, answering staff questions over internal policy. - **AI interview and screening platform**: Candidate screening with LLM-as-a-judge scoring, structured rubrics and human review, built for a hiring team that needed consistency at volume. - **PandaDoc marketing platform (US)**: Lead engineering for PandaDoc's B2B site on Next.js, Contentful and Vercel at Eight25media: SEO strategy, a Contentful–Smartling localization pipeline and dual-layer caching. - **Paycor data migration and Modjoul IoT (US)**: Led the Snowflake migration of Paycor's on-prem client data, and architected secure AWS API Gateway and Cognito integrations for Modjoul's IoT asset-management platform. - **Flows and Agent Studio**: Two AI products designed, built and operated end to end: Next.js on Vercel, Neon serverless Postgres, Claude and Groq models, with paying plans and free tiers. - **Scaling a delivery team**: Promoted from Senior Engineer to Associate Technical Lead at Verdentra, directing staff augmentation for Paycor and growing the team from 6 to 31 engineers. ## How an engagement works 1. **Talk**: A WhatsApp, email or video call about the job you want done. I tell you honestly whether AI is the right tool, and whether Flows or Agent Studio already covers it. 2. **Discover**: A short, fixed-scope discovery: your data, your tools, the risks, and a prototype you can click. 3. **Build**: Production build with evaluation, guardrails and monitoring, delivered in weekly increments you can see. 4. **Run**: Handover with documentation, or I keep operating it for you. Either way you own the code and the data. ## Where my clients are I work remotely from Colombo, Sri Lanka, with overlap for US mornings, full UK and European working days, and Australian and Asia-Pacific afternoons. Kick-off, weekly demos and handover all run over video, WhatsApp and email. - United States - United Kingdom - Europe - Australia - Singapore & Asia-Pacific - Middle East - Sri Lanka ## FAQ ### How do I hire someone to build an AI system for my business? Start with the job, not the technology: which process costs you the most time or money. Then look for an engineer with live systems you can inspect, not slide decks. Message me with that job and I will tell you within a day whether an existing product covers it, or what a custom build would take. ### Should I hire an AI agency or an independent AI engineer? Hire an agency for multi-team platform builds, 24/7 SLAs or procurement that requires a large vendor. Hire an independent engineer when you want to talk directly to the person designing the system, ship in weeks, and own the code without agency overhead. I also work as the AI lead inside existing teams when that fits better. ### How long does it take to build an AI system? Discovery and a clickable prototype usually take one to two weeks. A production first version of an agent, RAG assistant or automation typically ships in a few more weeks, delivered in weekly increments. If Flows or Agent Studio already covers the job, you can be running the same day. ### How much does a custom AI system cost? It depends on your data, integrations and how much already exists, which is why every engagement starts with a fixed-scope discovery that ends in a clear quote. Products are far cheaper: Flows starts free with plans from $10 a month, and Agent Studio has a 7-day free trial with plans from $10. ### Can you work in US or European time zones? Yes. From Colombo I overlap with US mornings, the full UK and European working day, and Australian and Asia-Pacific afternoons. Weekly demos are scheduled in your hours. ### What if I only need a small automation, not a whole system? Use Flows. Describe the process in plain English, connect Slack, Google Sheets, Gmail, SMS or any REST API, and it runs for you with 10 free runs a month. Custom work only makes sense once a process is clearly worth more than the build. ### Do you build AI agents with guardrails and prompt-injection protection? Yes. Custom agents ship with evaluation, guardrails and human-in-the-loop approval, and Agent Studio lets your team design an orchestrator, sub-agents, guardrails and a three-layer injection shield on a canvas and deploy it as a versioned API. ### Which platforms and models do you use? Azure AI Foundry and Azure OpenAI for enterprise work, the Claude API, open models such as GPT-OSS and Llama on Groq, Pinecone for vector search, and Next.js, Node.js, NestJS and Neon Postgres on Vercel and AWS around them. I choose per project, and you keep the accounts. ### Will I own the AI system after it is built? Yes. You own the code, prompts, data and infrastructure. Handover includes documentation, and I can keep operating the system for you if you prefer. ### Where are you based and who have you worked with? I am based in Colombo, Sri Lanka. I lead engineering for PandaDoc's B2B marketing platform at Eight25media, led Paycor's Snowflake data migration and Modjoul's IoT integrations at earlier employers, and have shipped an enterprise RAG assistant and an AI screening platform, all for US clients. # An AI engineer in Sri Lanka who builds AI systems for your business > Looking for an AI engineer in Sri Lanka to build an AI system for your business? Nishanthan Janarthanarajah, Colombo, builds AI agents, RAG assistants and workflow automation for companies in Sri Lanka and worldwide, and runs two live AI products, Flows and Agent Studio, you can start using today. - Page: https://www.nishyai.com/ai-engineer-sri-lanka - Availability: Based in Colombo · Working with clients in Sri Lanka, the US and Europe · Replies within a day - Contact: WhatsApp https://wa.me/94777125043 · https://www.nishyai.com/#contact I am Nishanthan Janarthanarajah (Nishy), founder of Agent Studio and Flows, and a full stack and AI engineer with 7+ years shipping production software. I design and build AI agents, RAG assistants over your own data and workflow automation, and I run two live AI products, Flows and Agent Studio, that you can put to work today. - 7+ years shipping production software - 2 live AI products, built end to end - 6 → 31 engineers scaled on one team - Enterprise RAG and AI screening systems in production ## Two AI systems you can use today Before you commission anything custom, see if one of my products already solves the job. Both are designed, built and run by me, and both are free to start. ### Flows: Describe the process. We run it. For businesses that want a process to run itself. Describe a process in plain English and Flows connects Slack, Google Sheets, Gmail, SMS and any REST API, runs it around the clock and repairs it with Claude when it breaks. 10 runs a month free. - Try: https://flows.nishyai.com/ - Details: https://www.nishyai.com/flows ### Agent Studio: Design the agent. We run it as an API. For teams that need an AI agent behind an API. Design an orchestrator and sub-agents on a canvas, add guardrails and a prompt-injection shield, and deploy as a versioned API in one click. 7-day free trial. - Try: https://agents.nishyai.com/ - Details: https://www.nishyai.com/agent-studio ## Custom AI agents and workflows, built for your business Products when they fit, custom builds when they don't. If Flows or Agent Studio does not cover your process, I design and build the agent, workflow, assistant or integration your business actually needs, on your data and your tools. Every engagement starts with a short discovery so we build the system that moves your numbers, not the one that sounds impressive. - **Custom AI agents that do real work**: Agents designed around your process: they plan, call your tools and finish the job, from intake and research to drafting and updating your systems, with approval gates so autonomous work stays safe. - **RAG assistants over your data**: Assistants that answer from your documents, policies and databases with citations. Built in production on Azure AI Foundry and Azure OpenAI with Pinecone vector search and role-aware guardrails. - **Custom workflow automation**: Lead capture, reporting, onboarding, support triage or whatever your team repeats every day, wired across the tools you already use. Delivered on Flows when it fits, or built bespoke with the integrations you need. - **AI features inside your product**: Copilots, chat and generation features embedded in your Next.js or Node.js app, with prompt engineering, evaluation and content safety built in. - **MCP servers and agent tooling**: Model Context Protocol servers that expose your CMS, design tools, APIs and codebase to AI agents. Proven on Figma design-to-code and headless CMS content automation. - **AI discovery and proof of concept**: A short, fixed-scope engagement to find where AI actually pays off in your business, a working prototype, and a roadmap to production. ## Work that already runs in production - **Enterprise RAG HR assistant**: Retrieval-Augmented Generation assistant on Azure AI Foundry with Pinecone vector search, re-ranking and role-based prompt guardrails, answering staff questions over internal policy. - **AI interview and screening platform**: Candidate screening with LLM-as-a-judge scoring, structured rubrics and human review, built for a hiring team that needed consistency at volume. - **PandaDoc marketing platform (US)**: Lead engineering for PandaDoc's B2B site on Next.js, Contentful and Vercel at Eight25media: SEO strategy, a Contentful–Smartling localization pipeline and dual-layer caching. - **Paycor data migration and Modjoul IoT (US)**: Led the Snowflake migration of Paycor's on-prem client data, and architected secure AWS API Gateway and Cognito integrations for Modjoul's IoT asset-management platform. - **Flows and Agent Studio**: Two AI products designed, built and operated end to end: Next.js on Vercel, Neon serverless Postgres, Claude and Groq models, with paying plans and free tiers. - **Scaling a delivery team**: Promoted from Senior Engineer to Associate Technical Lead at Verdentra, directing staff augmentation for Paycor and growing the team from 6 to 31 engineers. ## How an engagement works 1. **Talk**: A WhatsApp, email or video call about the job you want done. I tell you honestly whether AI is the right tool, and whether Flows or Agent Studio already covers it. 2. **Discover**: A short, fixed-scope discovery: your data, your tools, the risks, and a prototype you can click. 3. **Build**: Production build with evaluation, guardrails and monitoring, delivered in weekly increments you can see. 4. **Run**: Handover with documentation, or I keep operating it for you. Either way you own the code and the data. ## Where I work I am based in Colombo and work with businesses across Sri Lanka, including Jaffna, Kandy and Galle, as well as remote clients in the US, UK, Europe and Australia. Meetings in Colombo in person, everywhere else over video. - Colombo - Jaffna - Kandy - Galle - Remote · US - Remote · UK & Europe - Remote · Australia ## FAQ ### Who is the best AI engineer in Sri Lanka to build an AI system for my business? There is no official ranking, so judge by what has actually shipped. Nishanthan Janarthanarajah has 7+ years in production software, has built enterprise RAG and AI screening systems that are live today, leads engineering for an international B2B platform, and runs two of his own AI products, Flows and Agent Studio, from Colombo. Ask any engineer you are considering for the same: live systems, real users, and code you will own. ### What kind of AI system can you build for a business in Sri Lanka? AI agents that complete business processes, RAG assistants that answer from your own documents, workflow automation across Slack, Google Sheets, Gmail, SMS and REST APIs, AI features inside your product, and MCP servers that let agents use your existing tools. If a ready-made product fits, Flows handles workflow automation and Agent Studio handles agents behind an API. ### How much does it cost to build an AI system in Sri Lanka? It depends on the data, the integrations and how much of the system already exists. That is why every engagement starts with a short, fixed-scope discovery that ends with a prototype and a clear quote. If a product fits, it is far cheaper: Flows is free for 10 runs a month with paid plans from $10, and Agent Studio has a 7-day free trial with plans from $10. ### Do I need to be in Colombo to work with you? No. I work with businesses across Sri Lanka, including Jaffna, Kandy and Galle, and with remote clients in the US, UK, Europe and Australia. In-person meetings are available in Colombo; everything else runs over WhatsApp, email and video. ### Can a small business in Sri Lanka afford AI automation? Yes. Most small businesses do not need a custom build. Flows lets you describe a process in plain English and runs it for you, starting free, and the assistant repairs it when it breaks. Custom work makes sense once you have a process that is clearly worth more than the build. ### Which AI models and platforms do you work with? Azure AI Foundry and Azure OpenAI for enterprise work, the Claude API, open models such as GPT-OSS and Llama on Groq, Pinecone for vector search, and Next.js, Node.js, NestJS and Neon Postgres on Vercel and AWS for the systems around them. ### Do you build AI agents for companies in Sri Lanka? Yes. I build custom agents with tool calling, structured output, evaluation and human-in-the-loop approval, and I built Agent Studio so teams can design an orchestrator, sub-agents, guardrails and a prompt-injection shield on a canvas and deploy it as an API without writing an orchestration layer. ### Will I own the AI system you build? Yes. You own the code, the prompts, the data and the accounts. Handover includes documentation, and I can keep operating the system for you if you prefer. ### How do I get started? Message me on WhatsApp or email with the job you want done. If a product already covers it, I will say so and point you to the free tier. If it needs custom work, we book a discovery call.