Computer Science Undergraduate · MAKAUT · Kolkata, India
I build AI-powered full-stack applications that solve real problems. My work spans generative AI (LLMs, RAG, agents), computer vision (OCR, OpenCV), and modern web stacks (Next.js, TypeScript, FastAPI).
What I care about in a codebase: explicit over clever, working over perfect, and edge cases caught before they ship.
Languages
Frameworks & AI
Data & Tooling
Also working with: Gemini API, OpenAI API, retrieval-augmented generation, Pytesseract, WebSockets, and prompt engineering.
Voice-first assistant with real-time speech recognition, LLM reasoning, tool calling, and persistent memory.
Next.js TypeScript Express.js Gemini API WebSockets Deepgram ElevenLabs
- Agentic orchestration layer where the LLM selects tools (web search, calendar, tasks) from natural language rather than fixed intents
- Streaming audio pipeline with conversational memory across turns and a modular plugin interface for new tools
OCR and AI pipeline that automates medicine verification for donors and NGOs, cutting manual review out of the donation flow.
Next.js TypeScript MongoDB FastAPI OpenCV Pytesseract Gemini API
- Verification pipeline chains OCR extraction, LLM validation, expiry checks, and QR confirmation into a single pass
- Built both sides: computer vision service in Python and the React frontend against a MongoDB schema
Claim analysis grounded in live web evidence instead of model recall.
Next.js TypeScript Gemini API Serper API Tailwind CSS
- Retrieval-augmented generation over live search results, so every verdict cites retrieved sources rather than asserting from parameters
- Multilingual input with confidence scoring and structured citation output
B.Tech, Computer Science & Engineering — MAKAUT, West Bengal
Expected graduation 2028 · CGPA 8.80/10
Best reached at [email protected] or on LinkedIn. Work samples live on the portfolio.