Prajwal Kulkarni

I am a computer vision and machine learning engineer focused on building real-time, interpretable systems for medical and industrial applications. I combine strong academic training in electronics and communication with practical research and deployment experience.

My work includes developing a standalone application for automated evaluation of surgical training samples with integrated reports and Grad-CAM interpretability, and engineering agentic AI frameworks and eye-tracking integrations for surgical navigation.

I have built production-grade systems such as a multi-vehicle tracking service deployed with FastAPI, Docker, Hugging Face Spaces, and CI/CD, and a real-time posture analysis pipeline that achieved 88.2% accuracy using ST-GCN, MediaPipe, and SE attention blocks.

I pursue robust model behavior in high-stakes settings, exemplified by my research on calibration techniques for clinical imaging and large-scale weak-supervision pipelines that auto-labelled 58,000+ frames to create physics-aware datasets.