Last Updated
About
I follow my interests, and in the process do things, meet people, and make investments.
I’m a hands-on investor at my fund, ScOp VC, helping companies develop products and scale teams. I’m interested in vertical applications that leverage complex underlying data, like government, finance, and real estate.
Since 2024, I’ve been primarily interested in the big implications of AI and how it might impact society in ways we haven’t prepared for. I’ve gone back into research mode and collaborated on many academic publications, focused on the problem of reward hacking in agentic environments: its implications, detection, and mitigations. I’ve also concluded that benchmarking is essential both to advance AI and to do so safely. The speed of advancement and the voracious appetite for RL environments (environments and benchmarks are essentially the same thing) have made it increasingly difficult to maintain a high bar. As a core author of Terminal Bench, focused specifically on task quality, I’ve looked at hundreds of tasks and developed intuition for the limits of current approaches. I believe verifier quality is one of the most important problems in AI in 2026.
Prior to investing full time, I spent ten years at Graphiq, a startup we eventually sold to Amazon, where I remained for three years until 2020. At Graphiq we built one of the largest general knowledge graphs and developed natural language understanding and generation technology, which became Alexa’s pre-LLM Q&A engine. At Amazon I managed a team of 140, which gave me a lot of time to think about and experiment with management and organizational approaches to optimize product and software development and foster innovation. During the 2010s, my team developed nocode tools for “knowledge engineers,” a role we invented at Graphiq to organize the team around knowledge categories (sports, movies, geography). These interests have translated well into a world of LLMs, particularly in cases where the optimal solution involves hairy, domain-specific data.
My academic background is in Electrical Engineering and Mathematics, followed by a short stint as a PhD student in Financial Mathematics. I dropped out of my doctorate when a professor presented a proof for a stochastic process that resulted in infinite probability and refused to provide an intuitive explanation. That led me to conclude I was more of an engineer than a mathematician after all. Now I can claim a graduate education or brag about being a dropout, depending on the context.
Along the way, I’ve gone into side quests such as founding Unwrap.ai, a pre-ChatGPT bet on the future of NLP as a scalable unstructured analysis tool, and being CEO of HeyTutor, where we deliver tutoring services to public school children across the country.