About Traverse

Traverse is a research data lab building reinforcement learning environments for frontier AI labs. We focus on the non-deterministic, taste-dependent work that makes up most of the economy and that nobody else has figured out how to train models on. We work directly with the labs building the most capable models on earth as a thought partner. Backed by Y Combinator.

About the Role

As Head of Research, you will own the entire research function at Traverse, from setting the agenda and publishing papers to building benchmarks and leading the team that figures out how to train AI on the hardest problems in the world.

The core challenge is that most of the economy runs on work where "good" is subjective, feedback is delayed, and expertise takes decades to develop. Medicine, law, negotiation, engineering: none of these domains have neat reward signals, and figuring out how to encode what mastery looks like in these fields and turn that into RL environments that actually make models better is an open research problem. You'll be the person solving it.

You'll work directly with the frontier labs we partner with, understand their post-training pipelines, and co-design environments that reflect real hypotheses about how intelligence develops in a given domain. You'll publish novel work that establishes Traverse as a research institution, not just a data vendor.

We're looking for someone with serious research depth in ML and a track record of publishing at top venues, but we care just as much about your ability to reason about hard problems outside of ML. The best candidate probably has both: deep technical chops and genuine curiosity about domains they've never worked in.

In this role, you will

•⁠ ⁠Own Traverse's research output end-to-end - identify open problems, publish novel papers and create benchmarks that move the field forward

•⁠ ⁠Set the research direction for how we build and evaluate RL environments across verticals where verification is the hard problem

•⁠ ⁠Develop methodologies for encoding domain expertise into reward signals, evaluation criteria, and environment design

•⁠ ⁠Work directly with partner labs to understand their post-training pipelines and co-design environments that reflect real hypotheses about how intelligence develops

•⁠ ⁠Hire and lead a team of research scientists

•⁠ ⁠Represent Traverse as a serious research institution through publications, conferences, and direct relationships with the labs

Your background looks something like this

• PhD or equivalent depth of research experience in any rigorous field

• Published work in ML, in particular reinforcement learning, RLHF, reward modeling, or LLM post-training

• Track record of original work on hard, open-ended problems

• Ability to lead a team and set a research agenda

• Strong opinions and the ability to defend them

Bonus

• Deep expertise in a non-ML domain (medicine, law, finance, engineering) in addition to or instead of ML

• Prior work at a frontier AI lab and early-stage startup

Compensation

Competitive - Offers Equity, Bonuses