Member of Technical Staff - Research

What You'll Do

Push the Frontier of Judgment

• Research how frontier models reason through ambiguous, high-stakes business decisions - and where they fail

• Design novel methods for reasoning, evaluation, and calibration that go beyond standard benchmarks

• Translate open problems in reasoning, uncertainty, and multi-step decision-making into approaches we can test and ship

Build Evaluation That Matters

• Define what "good judgment" looks like for a model, and build the evaluation frameworks to measure it

• Design experiments that reveal real failure modes, not just leaderboard scores

• Turn findings into concrete recommendations for the product and applied teams

Partner with Founders

• Shape technical vision and roadmap alongside the founding team

• Bring outside research thinking into a company solving a problem few labs are focused on

Set the Bar

• Define what rigorous, applied research looks like in an AI-first organization

• Raise the bar for the team as it grows

Who You Are Must-Haves

• At least one undeniable signal of excellence - published research at a top venue, research role at a frontier lab, or a track record of novel technical contributions

• Deep understanding of how LLMs reason, fail, and can be evaluated - this isn't a theoretical interest, it's core to the job

• Strong fundamentals in ML, statistics, or NLP - we're too small for hand-holding on the technical side

• Comfortable moving from open-ended research question to a testable hypothesis quickly

Strong Signal

• Experience designing evaluation frameworks or benchmarks for reasoning, decision-making, or agentic systems

• Published or shipped work on uncertainty, calibration, multi-step reasoning, or LLM evaluation

• Has operated in a high-growth, high-ambiguity environment

• Thinks like an owner - big picture, not just your part

We're Not Looking For

• Researchers who need a clean, self-contained problem before starting - ambiguity is the job

• Pure benchmark-chasers - we care about judgment that holds up with real clients, not just leaderboard gains

• Researchers without product instincts - your work needs to change what we build and ship, not sit in a paper -- but we are open to publishing work!
  
  

Benefits & Perks

• Zero to Outcome: Seed stage means you shape the product, the culture, and the trajectory - not inherit them.

• Full Health Coverage: Medical, dental, and vision.