I build AI products end-to-end — agentic systems, RAG pipelines, and the full stack around them. Final-year B.S. (Honors) in Data Science & AI at IIT Guwahati ('27), shipping production LLM systems on the side.

Agentic AI Intern · Rowboat Labs (YC S24) (San Francisco — remote)

Shipping into an open-source AI coworker platform (15k+ ⭐). Built Skills — agents load packaged instruction sets and tools on demand, so new capabilities ship as installable packages instead of core-code changes. Also shipped ChatGPT OAuth (OAuth 2.0 + PKCE) so users can run agents on an existing Codex subscription instead of separate API keys. 30+ PRs merged into a TypeScript-first production codebase, both co-founders reviewing.

TypeScript Agent Orchestration MCP OAuth

Research Intern · RAIVN Lab, University of Washington (Remote)

Multi-agent web browsing: up to 4 vision-language agents drive isolated Playwright browsers in parallel through a screenshot→action loop, with best-of-N selection over a pluggable judge. Swappable policy layer with MolmoWeb-4B and Qwen-VL adapters, behind a grounding gate that catches coordinate bugs before they read as model error.

Vision-Language Models Multi-Agent Playwright LLM Evals

Founding Engineer Intern · ThinkSpace AI (Singapore) Multi-document research assistant as an Electron desktop app — sentence-window RAG with citations anchored to exact page coordinates, across PDF/DOCX, Google Drive and live web. Validated by a legal-advocate QA team against a ~100-question gold standard. LangChain RAG FastAPI Electron

AI Engineer Intern · Compeers AI (United States) Two market-research tools for a B2B intelligence platform: a 4-module pipeline (Google Custom Search → PDF/CSV parsing → SEC EDGAR 10-K → Google Trends → SWOT) and a Reddit audience profiler with NLP-based uniqueness scoring — replaced work that used to be manual. Python NLP Data Pipelines

🎯 CareerLift — AI career platform 5,000+ organic users. LLM agents parse your resume, semantically match it against live listings and 1,500+ IIT professor research roles, and return ranked roles with skill-gap analysis. The jobs pipeline refreshes 3,500+ Indian and international openings every 12–24 hours and pushes personalized alerts — no manual curation, zero paid spend.

🧠 Drona AI — autonomous interview agent 1,000+ organic users, 500+ in the first two weeks. Generates role-specific questions from an uploaded PDF, tunes difficulty from a rolling performance window, and streams a personalized feedback report. Re-architected from Streamlit to serverless Next.js on Vercel — dropped a ~500 MB ChromaDB stack for lightweight client-side scoring, same personalization at zero infra cost.

🔌 OpenCollab MCP — MCP server Open-source GitHub contribution matchmaker: skill-matched "good first issues", repo health scoring, PR plan generation. Zero-infra deploy via STDIO transport — installable through uvx in Claude Desktop and Cursor. MIT licensed.

AI / LLMs — Python · LangGraph · LangChain · MCP · RAG · FAISS/ChromaDB · PyTorch · HuggingFace · Vision-Language Models · LLM Evals Backend — FastAPI · Node.js · TypeScript · PostgreSQL · Supabase · Redis · MongoDB · Docker Frontend — React · Next.js · Tailwind · Vercel DSA — LeetCode 1710 (top 16.6%) · 1000+ problems

Demos are easy. Products people come back to are hard. I optimise for the second one — shipped over polished, end-to-end over one nice slice, real users over screenshots.

If you're a founder or team building with LLMs and want a hand, my inbox is open.

_{Open to AI engineering opportunities. Remote preferred, but flexible.}