Vals AI
Member of Technical Staff - Research, AI Safety
$130K–$185K
San Francisco · On-site · Staff · Full-time · Posted Oct 21, 2026
About the job
Designs and builds benchmarks and evaluation environments to identify misaligned behavior in frontier AI models and agents, and researches and communicates what those evaluations reveal.
What you'll do
- Develop and validate benchmarks for behaviors such as reward hacking, scheming, and deception, including in multi-agent systems.
- Analyze model behavior and collaborate with AI labs and enterprises on safety evaluation needs.
- Publish research findings and contribute to safety and evaluation communities.
What they ask for
- Substantial hands-on experience in safety evaluations, monitorability, scheming, deception, or multi-agent safety; safety research publication or high-quality write-ups are also cited.
- Deep ML/AI background, including neural networks and model training fundamentals.
- Excellent coding skills and ability to build complex evaluation environments, mostly in Python.
- Ability to communicate complex research to technical and non-technical audiences.
Nice to have
- Mechanistic interpretability experience.
- Contributions to open-source safety or evaluation tools and datasets.
Pay
$130K – $185K (base salary)
Tools
Python, Django, React, AWS, CDK
AI focus
Research and develop evaluations for AI safety and misaligned behavior in frontier models and agents.
Model-written summary · the original posting is authoritative