At SNR Edatas Private Limited, I develop and deploy computer vision applications for surveillance, safety, and inspection, from data preparation through production rollout. I build YOLO pipelines with PyTorch and OpenCV, then export models to ONNX for low-latency inference.
For election crowd monitoring, I built a real-time people-counting system for polling booths that reached 98% accuracy and supported about 1,000 cameras in a single election. It was deployed across multiple live state elections.
I also built a face-recognition entry/exit tracking system for a monitoring project across 100+ temples, with 90–95% accuracy, and an examination malpractice detection system deployed at exam venues. My work includes streaming camera feeds and sending alerts to operators.
I developed a Text-to-SQL assistant that lets operators ask text or voice questions about cameras, alerts, and video events without writing SQL. I’ve also created an auto-labeling tool that labels about 70% of objects on average, reducing the amount left for manual review.

