I build ML and data systems end to end, and I like the ones that have to notice when something is wrong: intrusion detection on live network traffic, entity resolution across millions of messy records, and RAG over real NGO data. Computer Engineering @ VIT Mumbai, CGPA 9.5/10.
| Project | What it does | Result |
|---|---|---|
| NetForecast · SIH 2026, team lead | LSTM world model on CIC-IDS2018, scored by Mahalanobis distance-from-normal | AUC 0.984 on unseen attacks (vs 0.672) |
| Amazon ML Challenge 2026 | Entity resolution across 3 sources in 72 hours | macro F0.5 0.712 → 0.882 at 7.6M records |
| GreenGuard · team lead | Plant-adoption platform for a Mumbai NGO; RAG with RRF over pgvector + FTS | 380+ commits, PostGIS matching |
| opensre | Open-source SRE tooling | EKS refactor, dedup fix, suite-wide test protocol |
| EVNet Sentinel | Reproduced an online IDS paper on CICEVSE2024 | Found timestamp leakage inflating results |
| MOSIP Decode 2026 | OpenID conformance runner for Inji Certify/Verify | Gates MOSIP releases |



