_{CSE (AI/ML) @ NIIT University · AI · Security · Full-stack}

• 🎓 Computer Science & Engineering (AI/ML) student at NIIT University, Neemrana
• 📸 Built ImpactProof, which turns NGO field photos into verified, donor-checkable evidence
• 🛡️ Built CodeSentry, an AST-aware RAG security auditor for unfamiliar codebases
• 🌐 Researched secure SDN routing with fuzzy logic (R&D project, supervised by Dr. Shweta Malwe)
• 🧱 Building full-stack products with Next.js, TypeScript, FastAPI and Node.js
• 🐳 Shipping with Docker, Docker Compose and Jenkins
• 💜 Made SoulCare, an emotional support chatbot for students living away from home
• 🦈 Merging pull requests on GitHub (Pull Shark achievement)

Python FastAPI Next.js PostgreSQL Cloudinary Gemini

Problem: Donors doubt impact photos: recycled images, wrong locations, stock and AI-generated pictures. Solution: Nine explainable checks (GPS, capture date, reuse, C2PA credentials, metadata, visual content, visible text, weather and more) give every photo a verdict: Corroborated, Needs review, Suspicious or Unverifiable. Uncertain cases go to a human review queue with an append-only audit log, and each claim gets a QR-coded public verification page. Built for: Code Cubicle 6.0 (Cloudinary problem statement).

Next.js Node.js FastAPI MongoDB ChromaDB Docker

Problem: Understanding and auditing an unfamiliar codebase is slow. Solution: Point it at a public GitHub repo and ask questions. Hybrid retrieval (embeddings + BM25L) with cross-encoder reranking surfaces answers and security findings with highlighted citations, and each repo is isolated so context never leaks between codebases.

Python Mininet Ryu OpenFlow

Problem: OSPF and RIP will happily route through a fast node that has been compromised. Solution: A 6-step security pipeline feeds a trust multiplier into a fuzzy routing score, so a compromised node scores zero and is never selected.

TypeScript Next.js Tailwind

Problem: AI agents need to pay for things, but unrestricted wallet access is dangerous. Solution: A product concept for bounded financial autonomy: agents work within user-defined budgets, permissions and escrow. Simulation only, no real funds move.

Next.js TypeScript Vercel

Problem: Students far from home sometimes just need someone to talk to. Solution: An empathetic chatbot for daily check-ins, with weekly reports and an insights dashboard that help users reflect on their mood over time.

TypeScript

Open-source toolkit for connecting your users to their apps.

Languages: Python · TypeScript · JavaScript Frontend: React · Next.js · Tailwind CSS Backend: Node.js · Express · FastAPI Data & AI: MongoDB · PostgreSQL · ChromaDB · RAG · Gemini DevOps: Docker · Jenkins · Linux · Git · Vercel · Render Networking: Mininet · Ryu · OpenFlow

_{Thanks for stopping by. If something here interests you, drop a ⭐}