I’m a principal scientist at NVIDIA, where I work on foundation models across Nemotron LLMs and Cosmos world foundation models. I also publish papers, contribute deep learning models to the community, and serve in various roles at NeurIPS, ICLR, and CVPR.

From 2020 to 2022, I was a research scientist at Google Research, where I co-created FILM interpolation that powers user-facing features in Google Photos, Pixel and featured in Google I/O; and TryOnDiffusion that launched on Google Search and featured in 23 of Google’s biggest moments in 2023.

Previously, I was a research scientist at NVIDIA Applied Deep Learning Research, led by Bryan Catanzaro. I co-founded and led the real-time graphics technology that exploits the temporal domain to increase the performance of GPUs in video games, which culminated in the creation of DLSS 3.0. I’ve also pulished on topics spanning semantic segmentation, video prediction, image inpainting, and frame interpolation, that were featured in Fortune, Forbes, and FastCompany.

Prior to that, I was the lead inventor and researcher of the team that launched the Siemens FastSpine technology that automatically traces, detects and numbers the human spine in 3D CT/MRI images.

I received my PhD in Electrical Engineering from Vanderbilt University. My thesis was on image processing algorithms for image-guided surgery techniques. I hold a MS degree in Computer Vision and Robotics, received from Heriot-Watt Univerisity.