With RUEXIS AI, I architected a multi-agent compiler that turns academic papers into runnable PyTorch repositories. Across 48 evaluated papers, it achieved 100% AST validity and passed 203 regression tests.
For my M.Tech thesis at Pondicherry University, I developed E-SGCD, a semantic-guided framework for satellite change detection. It reached 90.97% F1 on the LEVIR-CD benchmark, with an edge-deployable design using Swin-Tiny and RemoteCLIP.
For DermSemble, my B.Tech capstone, I created a dual-stream deep learning system for clinical dermatology triage and served it as a FastAPI microservice. Its ONNX Runtime INT8 model delivered 11.3x edge compression and 7.2 ms GPU latency.
I also published research in the Journal of Electrical Systems and Information Technology and have a Medi-Assist AI paper accepted for presentation and publication. My projects span agentic AI, computer vision, RAG, and full-stack systems; I’ve also worked as a Full Stack Web Developer Intern at NullClass and as a freelance software and full-stack engineer.
