Come work at the limits of model behavior

Researchers and engineers formerly at Meta Superintelligence Labs, Amazon AGI, Apple and Google.

We build worlds for models to learn in and study where they fail. Come join us.

Open roles ↓How we work ↓
The Patronus AI team in Lyon

Open roles

Department

Roles

Role title

Department

San Francisco

Apply ↗

How we work

Solve hard problems at the frontier of AI research.

Relentless Optimism

We meet complexity with confidence, believing that effort, creativity, and persistence create breakthroughs.

Unlimited Energy

We bring momentum to everything we do, moving fast while maintaining the stamina to go the distance.

Intellectual Flexibility

We thrive in uncertainty, embrace complexity, and evolve continuously.

Patronus AI engineers working together in Lyon
A Patronus AI researcher at work during a day in Lyon

Questions about working at Patronus AI

What does Patronus AI work on?

Patronus AI is a behavioral research lab for AI that studies where models fail and builds training data and simulated environments to improve them. Its researchers and engineers build RL environments for AI agents, run reinforcement learning and post-training experiments, and publish benchmarks across domains such as coding, computer use, knowledge work and more.

Is Patronus AI hiring for RL environment roles?

Yes, Patronus AI hires researchers and engineers to build RL environments for training and evaluating AI agents. Its Member of Technical Staff roles in research and engineering cover environment and reward design, agent infrastructure, reinforcement learning and post-training methods such as GRPO and SFT. Current openings are listed on patronus.ai/join, pulled live from the company job board.

Can I work remotely at Patronus AI, and where is the office?

Patronus AI’s San Francisco research and engineering roles are in office five days a week because the team works closely together. Remote work is possible for specific circumstances and candidates and is agreed case by case. Domain expert contracts are fully remote and asynchronous. Experts work with the research team to design the tasks and grading behind RL environments.

Do researchers at Patronus AI publish?

Yes, researchers at Patronus AI publish their work. Papers from 2026 include main-track work at NeurIPS and ICML, such as a language model that stands in for an environment during agentic RL and a study of reward exploits in code RL, plus workshop papers at NeurIPS, ACL and COLM. The full list is on patronus.ai/research.

What is the research culture like at Patronus AI?

Patronus AI keeps research and engineering under one roof, so researchers own projects end to end, from framing an open problem to experiments, RL environments and production systems. The team works on reward design, agent behavior, reward hacking and scalable oversight, writes production code alongside engineers, and gets together at recurring research events in San Francisco and at global offsites.

Can domain experts work with Patronus AI on contract?

Yes, Patronus AI hires expert practitioners as remote contractors to design realistic tasks that today’s AI models get wrong. Experts write the request, solve it and define how the answer is graded, and that work becomes training and evaluation data for frontier models. No AI background is needed. Fields include finance, law, healthcare and more.

Who founded Patronus AI?

Patronus AI was founded in 2023 by Anand Kannappan, its CEO, and Rebecca Qian, its CTO. Before Patronus, Anand led ML explainability and experimentation work at Meta Reality Labs, and Rebecca published research at FAIR (Meta AI). Today the team includes researchers and engineers formerly at Meta Superintelligence Labs, Amazon AGI, Apple and Google.

Is Patronus AI a good place to work for AI researchers?

Patronus AI suits researchers who want to work on RL environments, agent behavior and post-training at a lab that publishes. Its team comes from Meta Superintelligence Labs, Amazon AGI, Apple and Google, its papers appear at NeurIPS and ICML, and it raised a $50M Series B led by Greenfield Partners in June 2026.