Our Mission
RewardGuard exists to make reinforcement learning safer and more reliable. We believe that detecting and correcting reward misalignment should be as easy as running a linter. We're building the tools to make that a reality — and we're looking for people who care deeply about getting AI right.
Our Values
Rigor First
We don't ship guesses. Everything we build is grounded in research and validated on real training runs.
Open by Default
Our core analysis tools are open source. We believe the safety community benefits from shared foundations.
Move with Purpose
We're a small team that ships. No unnecessary meetings, no bureaucracy — just focused, high-impact work.
Honest Communication
We say what we think, give direct feedback, and assume good intent. No politics.
No open positions right now
We're a small, early-stage team. We don't have open roles at the moment, but we're always interested in hearing from exceptional people who care about AI alignment.
Send us your resume →What we offer
Fully remote — work from anywhere
Competitive salary + equity
Unlimited PTO with a 3-week minimum
Top-of-the-line equipment budget
Learning budget for conferences and courses
Work on genuinely hard, important problems