At Amzur Infotech Pvt. Ltd., I lead architecture and development for Agentic-AI projects. I delivered a governed legacy modernization framework that orchestrates AI agents through migration stages and verified a Rails-to-Nest.JS and React18 migration.
At ZestIoT Pvt. Ltd., I led work on Predicton, a vision-based conveyor belt health monitoring system, integrating PLC functionality and defect classification that achieved 98% average precision. I also developed aviation analytics, including a flight turnaround estimation algorithm with 94% accuracy.
Earlier, I built deep learning and computer vision systems for industrial inspection, including welding defect detection with 98% accuracy on unseen defects and a U-Net model for wooden-surface damage detection. In personal projects, I implemented an LSTM block from scratch and a residual U-Net that achieved 96% IoU on the DAGM dataset.
