At DRDO Centre for AI and Robotics (CAIR), I built an NLP pipeline to interpret noisy free text, using T5 correction, entity retrieval, and semantic ranking across a 600,000+ row PostgreSQL dataset. It reached 79.4% accuracy and 88.18% F1 on 2,000 test records, was deployed as a REST API, and was published at IEEE CICT 2025.
At iAgami Technologies, I built the data layer for NutriPersonalised, a personalised health and nutrition product, with a RAG pipeline over PostgreSQL and AWS storage. I also developed and deployed CNN models for hair and scalp analysis, leading data curation and pixel-level annotation.
At L&T Technology Services (LTTS), I built a reasoning engine that discovers exact rules from numerical and tabular data using symbolic regression and a genetic algorithm. Its two-stage validation pipeline cut verification work by about 50%, with performance benchmarked against LLM baselines.
I won the Mercury Award at the L&T Global Open AI Hackathon 2026 for MedAI: Agent Control Tower, a 16-agent system that runs a patient journey from emergency intake to discharge. My other healthcare AI projects include MRI diagnosis, chest X-ray analysis, and speech-enabled triage.

