Alex Shaw

These days I spend my time building Harbor, an open-source framework for building RL environments and agent evaluations.

I created Harbor based on my experience building Terminal-Bench, a popular agent evaluation I co-created at Laude Institute.

Before working at Laude, I worked on user modeling and causal inference at Google and studied math and computer science at BYU.

I currently live the in Bay Area and am eager to meet folks who are interested in pushing the boundaries of what we can simulate with RL environments and how we can make it easy to run billions of simulations, extract useful reward signals, detect agent failure modes, and post-train custom models.

If you’re interested in joining the Harbor team to build a future where core IP takes the form of RL environments in a world of continually learning agents and billions of custom models, please reach out :)