Summary
Gaurav Goswami is a Machine Learning Engineer with 8 years of hands-on experience building and optimizing deep learning models for computer vision, specializing in image/video denoising, low-light enhancement, and on-device inference. He led deployment of high-resolution imaging pipelines for multiple Samsung Galaxy flagship devices, driving quantization-aware training, model compression, and NPU/DSA-targeted optimizations to meet real-time mobile constraints. At Red Hat he now contributes to open-source AI infrastructure and scalable training/inference tooling, working with PyTorch ecosystem projects and vLLM to enable high-throughput model serving. His open-source work includes re-parameterizable denoising networks, a PyTorch quantization framework, and multiple object detection implementations—demonstrating a strong practice of turning research ideas into production-ready code. Comfortable across embedded SoCs, Jetson platforms, and server-scale systems, he blends low-level optimization with large-scale model orchestration. An intriguing through-line in his work is repeatedly shrinking state-of-the-art models to run with minimal accuracy loss on constrained hardware, from Jetson Nano to flagship phones.
7 years of coding experience
3 years of employment as a software developer
Bachelor of Engineering, Computer Engineering, Bachelor of Engineering, Computer Engineering at Mahatma Jyotiba Phule Rohilkhand University (MJPRU), Bareilly
English, Hindi