• 🎓 2nd-year B.Tech in AI & Robotics at UEM Kolkata (GPA: 8.19)
• 🔭 Currently building multi-agent AI platforms, RAG pipelines, and geospatial AI systems
• 💡 Passionate about solving real-world problems with LLMs, vector databases, and scalable APIs
• 🧩 Solved 100+ LeetCode problems in C++ — Arrays, Trees, Recursion & more
• 📜 Certified in Machine Learning Specialization by DeepLearning.AI (Andrew Ng)
• 📍 Based in Kolkata, West Bengal

Orchestrates 3 specialized AI agents (Research, Analysis, Coding) from a single natural language goal using a custom planner with intent detection.

• 🔍 Dual Tavily web searches with relevance scoring + ChromaDB cross-session memory
• 💡 Returns personalized startup ideas with market size, competitors & risk assessment
• 🔧 FastAPI REST layer with Pydantic v2, in-memory task store + n8n workflow automation

FastAPI GPT-4o ChromaDB Tavily Pydantic v2 n8n

Full-stack disaster relief platform connecting affected users to the nearest NGO using PostGIS spatial queries within a 200km radius.

• 🎙️ Multi-modal ingestion: text, Whisper STT, Tesseract OCR → GPT-4o structuring agent
• ✅ Two-stage AI validation with confidence thresholds + live stock checks
• 🗺️ NGO admin dashboard with severity-colored live map + JWT-authenticated inventory CRUD
• 🚀 Deployed on Railway + Vercel + Neon with Sundarbans flood scenario seed data

FastAPI PostgreSQL PostGIS GPT-4o Whisper Next.js 14 Redis Railway

RAG pipeline ingesting Tesla's 10-K PDF, generating embeddings via HuggingFace, stored in Pinecone, answered by Google Gemini.

Python LangChain Google Gemini Pinecone HuggingFace

⚡ "Build systems that think, scale, and solve." ⚡