About Sufian Md. Waliullah. AI engineer in Bangladesh.
Sufian Md. Waliullah is an AI engineer based in Bangladesh (GMT+6). He builds AI-powered web applications, with LLM features, retrieval-augmented generation (RAG) and agents, on a full-stack base of Next.js, React, Node.js and MongoDB, and works remotely with clients and teams.
By Sufian Md. Waliullah · Updated
In short. The facts people usually ask for first.
- Name
- Sufian Md. Waliullah
- Role
- AI engineer (LLM apps, RAG, agents) on a full-stack JavaScript base
- Based in
- Bangladesh, working remotely
- Time zone
- GMT+6 (Asia/Dhaka)
- Core stack
- OpenAI API, LangChain, Vercel AI SDK, Next.js, React, TypeScript, Node.js, MongoDB, PostgreSQL
- Open to
- Internships, freelance projects and collaborations
- Contact
- waliullahsufian@gmail.com

Background. Why full stack.
I like owning a feature from the first schema sketch to the deploy. Working across the whole stack means fewer handoffs, and it means I understand why the API looks the way it does when I'm building the screen that calls it.
I keep code typed, readable and small enough to change. I deploy early so there's always a working link to click, and I write down decisions so the next person (often me) knows why.
Most of my work is in JavaScript and TypeScript from end to end. I pick MongoDB or PostgreSQL based on the shape of the data, keep APIs small and documented, and put a working preview link in front of people as early as possible.
Skills. What I use, and what for.
AI
- OpenAI APIGPT models, tool calling, structured output
- LangChainRAG pipelines and agents
- Vercel AI SDKStreaming chat and tool calls in React
- Hugging FaceOpen models and embeddings
Frontend
- ReactComponents, hooks, state
- Next.jsApp Router, server rendering, routing
- TypeScriptTyped props, APIs and data models
- Tailwind CSSResponsive, utility-first styling
Backend
- Node.jsRuntime for APIs and scripts
- ExpressRouting, middleware, validation
- PythonFastAPI services and data scripts
- REST APIsResource design, status codes, errors
Data
- MongoDBDocument modeling and indexes
- PostgreSQLRelational data and joins
- pgvectorEmbeddings and similarity search
- PrismaType-safe queries and migrations
How I work. Four steps, repeated in small loops.
- Step 01
Understand
We talk through the problem, who uses it and what done looks like. I write it down so we agree on scope before any code.
- Step 02
Model the data
I sketch the collections or tables and the API routes first. Getting the schema right early saves rewrites later.
- Step 03
Build a thin slice
One feature working end to end and deployed to a preview link you can click, usually within the first few days.
- Step 04
Ship and iterate
Small releases with short notes on what changed. Feedback goes straight into the next slice.