Summary
Bas Van Opheusden is a research-driven AI scientist and Member of Technical Staff at OpenAI with 10+ years of interdisciplinary experience at Princeton, NYU, Meta Reality Labs and Imbue, where he helped train and fine-tune large language models and built in-house evaluation pipelines. He combines a PhD in neuroscience and a Cum Laude background in physics and mathematics to study how human cognition and social structure shape learning, planning, and expertise, publishing in Science, PNAS, Nature Human Behavior and top AI venues. At Imbue he led large-scale human annotation efforts (500,000+ judgments) and linked ambiguity in prompts to LLM reasoning failures, while at Princeton he analyzed 3.3 billion chess games and ran the largest-ever interactive Mechanical Turk study of algorithm learning. Skilled at bridging experimental design, computational modeling and production ML systems, he focuses on using cognitive science to interpret and improve model internals and recommendations. Based in San Francisco, he brings a rare blend of rigorous empirical research and hands-on model engineering that surfaces non-obvious links between human decision strategies and model behavior.
10 years of coding experience
7 years of employment as a software developer
Doctor of Philosophy - PhD, Neuroscience, Doctor of Philosophy - PhD, Neuroscience at New York University
Master of Science - MS, Physics, Cum Laude, Master of Science - MS, Physics, Cum Laude at Leiden University
Dutch, English