Open roles

AI Researcher

Research · Full time


About the role

About the role

We're building at the frontier of the agentic web, a field being invented in real time. You'll help us understand it faster than anyone else by exploring how AI agents actually behave and turning what we learn into new standards we share with the world. You'll also help establish us as a top-tier research lab known for the depth and rigor of our work.

This isn't a standard research role. It sits between research and engineering, and it's about trustworthy knowledge put to work, not papers for their own sake. Think of it as a judgment role with a high technical floor. You'll decide which questions matter, design the experiments that answer them, build the tooling to run them, and tell us clearly what's true and what to do about it.

What you'll do

  • Turn fuzzy questions about agent behavior into clean, falsifiable experiments

  • Build and run your own experiment tooling end-to-end

  • Design evals and benchmarks that hold up statistically

  • Distill messy investigations into clear conclusions that drive agent-engine and research decisions

  • Publish work that builds our reputation as a lab and helps shape agentic-web standards

What you bring

You'll need all of the following.

  • Rigorous experimental design. You design clean experiments, spot confounders before they burn a week, and hold your work to a standard that survives external scrutiny.

  • LLM internals, grounded in ML and statistics. You understand how models work under the hood, so you can explain agent behavior, not just observe it.

  • Evals and benchmarking. You build evals that hold up statistically and know when a metric is lying.

  • Self-sufficient engineering. You build your own tooling and stay productive in agentic harnesses that break and change weekly.

  • Clear conclusions. You can say what's true, how certain it is, and what we should do, whether in an internal decision or a published piece.

  • Energy for the unknown. Working without a playbook or prior art energizes you. You build the method that doesn't exist yet.

  • Calibrated speed. You ship good-enough findings fast and kill dead ends quickly, but slow down on the claims a decision rests on.

What will make you stand out

  • Agentic protocols. You know the emerging agentic-web standards, their tradeoffs and where they're heading, and you move fast when a new one drops.

  • Web fundamentals. You know how the web works beneath the agentic layer, including HTTP, crawling, rendering, browsers and discovery.

  • AI search. You understand how LLMs and agentic search discover, rank and cite web content, and the shift from SEO to AEO it's driving. You already think about how a brand gets surfaced in an AI answer.

Background and qualifications

  • M.Sc. or Ph.D. in ML, NLP, statistics or a related field, or equivalent research experience

  • 5+ years in ML, AI research, or agent or systems engineering

  • Strong coding in Python / TypeScript with modern tooling

  • Hands-on experience building, evaluating or researching LLMs or agents

Experience at an early-stage startup, and comfort with the ambiguity that comes with it, is a bonus.


Apply

This site is protected by reCAPTCHA and the Google Privacy Policy and Terms of Service apply.