Pragyan Das

I've built safety-monitoring systems for Hyundai vehicle test tracks and a Schneider Electric warehouse in Singapore, taking work from GigE camera bring-up through YOLO/TensorRT inference, tracking, rule logic, FastAPI services, and operator tooling.

At Schneider Electric Singapore, I led the Critical-severity forklift exclusion-zone rule and reduced zone-centring error from 1.20 m to 0.17 m. I also built a hardware-independent simulation harness that uncovered geometry defects missed by existing metrics, including a camera-tilt error that could miss safety violations.

For Hyundai, I diagnosed ByteTrack instability, dead cameras, frame overwrites, and misrouted analytics workers, helping deliver roughly 2.2× throughput. I trained a seatbelt-detection YOLO model and designed an iRayple GigE Vision camera pipeline with software auto-gain using YOLO face boxes.

I prefer building to theory, from local voice assistants and transformers to bipedal robots, visual tracking, and electronics. I'm hands-on with servos, IMUs, 3D printing, soldering, and production ML systems.