Aadit Agrawal

GPS telemetry at 500K+ datapoints; dashcam CV with OpenCV + YOLO on 12,000+ labelled samples. Built a federated learning pipeline for privacy-preserving route optimisation across distributed edge nodes.

Authored long-form technical references on: TensorRT-LLM/Triton production serving, AMD ROCm, Apple MLX, RLHF/DPO, MoE architectures, edge AI, LLM fine-tuning (LoRA/QLoRA), HPC optimization, and VLMs.