B.S. + M.S. Computer Science

Building LLM steganography eval benchmark for AI safety and mechanistic interpretability system using SAEs on frontier model activations (Claude/GPT-4) to decompose 12k-dim representations into interpretable features. Collaborating with 3 researchers on alignment methods; preparing submissions to ICML and NeurIPS workshops.

Built Kapok, a cross-platform disaster relief coordination app (Flutter/Dart) with Firebase real-time backend and Hive offline DB for connectivity-limited environments. Integrated Mapbox API for real-time rescue visualization; agile team of 4 engineers.

Taught 210+ students Java, Python, and core CS concepts — sorting, recursion, data structures — across multiple semesters. Individual tutoring, grading, and weekly office hours.

Led ML project (1500+ LOC, 300k+ observations) predicting tropical tree phenology with 92% accuracy; improved model 78% through feature engineering. Presented novel phenological phase estimation model to SeasonWatch and 140+ students across India.

Deployed YOLOv8 pipeline for real-time road damage classification at 72ms / 21 FPS, trained on 52k images achieving F1 0.67. Managed datasets on AWS S3 with automated preprocessing via OpenCV and Albumentations.

Developed confidential ML models for stock time-series prediction using Keras. Technical analysis on equity, fixed income, and derivatives using Refinitiv Eikon and Pandas.

Built REST API for a college-based food delivery startup using NodeJS and ExpressJS. Implemented user data organization and retrieval system on Firebase Cloud.

Built a platform that lets users deploy networks of collaborating AI agents from a single prompt. Engineered Go/gRPC backend orchestration, Redis/PostgreSQL/vector DB persistent memory, and Kubernetes auto-scaling for multi-tenant enterprise. Recruited GAP's CTO as advisor; acquired first enterprise customer running production pilots.