FULL-STACK / AI-NATIVE / DEVELOPER
I build software that thinks for itself.
I work across the stack, frontend to backend to the model layer, and I'd rather ship something useful than something that just looks impressive in a demo.
A builder who treats models as products.
I'm a full-stack developer who lives where application engineering meets applied AI: training models, wiring up retrieval pipelines, and building the interfaces that make all of it usable by actual people.
My favourite work connects the whole stack. A frontend someone can actually use, a backend that holds up under real conditions, and an AI layer (retrieval, agents, or a fine-tuned model, whatever the problem calls for) that earns its place instead of getting bolted on because it's trendy.
I care about clean abstractions, reproducible pipelines, and writing code I can still make sense of six months later without wanting to rewrite the whole thing. Right now that means going deeper on agentic workflows, retrieval-augmented generation, and the Model Context Protocol.
Outside of finished projects, I like taking things apart just to see how they work, then putting them back together slightly better than before.
Where full-stack meets AI-native.
A tighter view of the tools I actually reach for when building products, AI workflows, and infrastructure.
Product UI
Interfaces and app shells.
- React
- Next.js
- TypeScript
- Tailwind CSS
Backend
APIs, services, and orchestration.
- Node.js
- Python
- FastAPI
- Express
AI Systems
Models, retrieval, and agent workflows.
- Anthropic
- LangChain
- LangGraph
- PyTorch
Data & Cloud
Storage, deployment, and runtime.
- Docker
- Redis
- AWS
- MongoDB
Work I keep coming back to.
A mix of AI-native and full-stack builds. Real prototypes, not polished commercial products, so expect a few rough edges.
Drag it, scroll it, or just click a project to open it
LookMeNot
Published Firefox add-onRead Langfuse traces, tool calls, and agent actions as a clean conversation pane in Firefox.
- · Conversation-style trace reading, not raw nested JSON
- · Search across messages, tool names, and tool inputs
- · Everything processes locally in your browser, nothing sent to a server
Problem-solving, in the open.
A live look at LeetCode activity, synced directly from a public profile. No login required.
sunilkumarweb47
Awaiting sync, showing fallback data
Where I show up professionally.
Selected LinkedIn presence, built from public profile information. No scraping, nothing fabricated.
Sunil Kumar Pradhan
Full-Stack AI-Native Developer
Bhubaneswar, India
Full-stack developer working where application engineering meets applied AI: training models, wiring up retrieval pipelines, and shipping the interfaces that make all of it useful.
Where I've shipped work and studied.
Let's build something meaningful.
Have an interesting project, opportunity, or technical problem? I'm always up for a good conversation about full-stack engineering, AI-native products, and developer experiences worth using.
The contact form isn't connected to a submission endpoint yet, so the fastest way to reach me is directly by email.
sunilkumarweb47@gmail.com