🎓 B.Tech IT @ VJTI Mumbai — Batch of 2029, minor in Entrepreneurship & Startups

🔍 The pattern in everything I build: something is hiding in the data — a fraud, a leak, a broken dependency, a species, a storm — and I build the thing that finds it

🤖 Into: multi-agent systems, ML that survives contact with real data, and tools that explain why

🏁 Hackathon regular — 24-hour builds are my favourite way to ship

🚩 CTF player — picoCTF, IIT Madras CTF (RSA, SQLi, JWT, stego, forensics)

♟️ Chess Coordinator @ Enthusia, VJTI

🚀 Endgame: build my own startup

• 📈 AlgoTrade ML — regime-aware ML trading system, live paper-trading on Alpaca since Mar 2026 (+24.7%, walk-forward Sharpe 1.18)
• 🌱 Gearing up for GSoC in open-source finance
• 🧠 Going deeper on LangGraph agents & graph-based fraud detection

Agents · fraud detection · ML for weather & biology · startups · chess

_{ELEVATE 1.0 · Top 30 of 900+ teams · built in 24h}

AI audit platform that checks 100% of invoices with 16 fraud detectors, graph-based circular-trading detection and an MCP server with 53 role-restricted tools. A ~40-hour audit runs in 60 seconds.

code · live demo ↗

_{🥈 Runners-up, Track 2 · Hacktoberfest 2026, K. J. Somaiya}

Multi-agent system that finds and safely fixes outdated Python dependencies. Fixed 18/18 failing checks with 0 regressions on local LLMs ($0 cost) — cutting 2–4 hours of debugging to 10 minutes.

code

_{Smart India Hackathon 2026 · eDNA biodiversity pipeline}

Identifies marine species from DNA fragments found in seawater and computes biodiversity indices. 98.85% accuracy with 192-D embeddings, and recovered 97.59% of reads other pipelines throw away.

code · live demo ↗

_{Research project · co-authored paper · team lead}

Forecasts North Indian Ocean cyclone intensity and track from 39 years of data. Random Forest hit 5.96 kt RMSE (R² 0.947); a hybrid Markov + physics model beat the persistence baseline at 72 hours.

code