Summary
Steven Schwarcz is a research scientist with 11 years of experience specializing in machine learning for vision and language, currently working on post-training research for Gemini at Google DeepMind and improving factuality and instruction-following in large language models. He blends rigorous academic training (PhD, University of Maryland) with product-focused applied science from roles at Amazon, where he redesigned ML pipelines and developed novel transformer architectures that dramatically improved precision and recall for tracking and action recognition. His background includes top-ranked performance on government-funded action recognition challenges, multiple WACV publications, and internships at Google Brain and StreetSmart where he advanced real-time tracking, pose estimation, and unsupervised domain adaptation. Equally comfortable designing new architectures and engineering practical tooling and metrics, he has a track record of turning prototype research into transformative system improvements. Based in New York, he brings a rare combination of real-time video expertise and LLM-focused post-training research to bridge perception and language systems.
11 years of coding experience
2 years of employment as a software developer
Doctor of Philosophy (PhD) Computer Science, Doctor of Philosophy (PhD) Computer Science at University of Maryland
Bachelors of Science Computer Science, Bachelors of Science Computer Science at The College of New Jersey