At the AI in Healthcare Lab at Indian Institute of Technology - Delhi, I fine-tuned Qwen3.5 on ultrasound image-text pairs and developed vision model training and evaluation pipelines. I also worked on breast cancer lesion detection and source-free domain adaptation for adapting models to an unlabeled Indian dataset.
I optimized PyTorch models for Jetson edge devices by converting them to ONNX and TensorRT and profiling inference trade-offs. At CommandHQ, I engineered and deployed a GCP AI backend that cut costs by 30% while increasing capacity tenfold, and built a Gemini-powered tool serving Army veterans transitioning to civilian life.

