At ZestIoT, I built and improved real-time computer vision systems for airport turnaround and baggage-handling operations. My work included a passenger-counting pipeline using YOLOv7 and ByteTrack to distinguish crew from passengers.
I fine-tuned and tested YOLOv12-s and RT-DETRv2, comparing them with YOLOv7 using ONNX and TensorRT. I also developed optical-flow-based baggage loading and unloading detection at airport sites in Lyon.
At IIT Bombay, I built a monkey detection system for Raspberry Pi-compatible hardware, selecting EfficientDet-D0 and applying quantization to achieve over 6 FPS on a single thread. I also developed an active-learning demo and an Aadhaar masking pipeline.
In personal projects, I built an OpenCV tracking system integrated with Arduino UNO for a robot dog, improving tracking accuracy and reducing servo jitter. I also trained a DCGAN for medical image inpainting and built a Streamlit app to create masks and view restored images.
