At HIOKI E.E. Corporation, I developed a tile-based OpenCV pipeline to extract electrical nets from PCB images. I also designed a spatial-indexing engine for automated insulation and clearance verification.
On the MFAD deepfake-detection project, I collaborated on a forensic pipeline that combined seven agents and Bayesian score fusion. The system achieved 99.0% ROC AUC across eight GAN-based datasets, and I engineered its geometry agent.
I built an intent-aware hybrid retrieval engine combining lexical and semantic search with a Neo4j knowledge graph; the project won first place at Hack60 by HCLTech. I also contributed to Beax, an IIT Mandi utility app with more than 3,000 users.
