Pentagon GamesWhat you'll do
- Build eval suites for agents and POCA: accuracy, refusals, tool misuse, prompt injection
- Red-team agents that touch accounts, payments and community spaces
- Define release gates: nothing ships to players without passing them
- Investigate incidents and turn them into tests
What you'll bring
- Experience evaluating or red-teaming LLM systems
- A security mindset: you think about how a system gets abused
- Python and test engineering
- Precise, evidence-first reporting
Nice to have
- Background in trust & safety or application security
How you'll work with AI
You will work alongside our own AI agents every day: they draft, test, research and ship with you, and part of your job is making them better.
Where and how
Remote · Taiwan · Hong Kong · Singapore · Dubai (housing benefit) · Toronto. Full-time. We work remote-first with offices in Taiwan, Hong Kong, Singapore, Dubai and Toronto; Dubai roles include a housing benefit.
How to apply
Apply on members.pentagon.games. You'll sign in or create a free Pentagon account first, so your application is saved to it and you can follow or withdraw it. Every application asks one question we care about most: what have you built with AI agents? Show us.