AI Safety Benchmark Research Engineer

Vals AI, Inc.; San Francisco (CA); On-site; USD 180,000 - 240,000; Full time

Benefits offered by this job

Relocation and transportation support

Health/dental insurance coverage

Lunch and dinner provided

Unlimited PTO

$1,500 housing stipend

Collaborate with leading AI labs

Publish and present your work

Job summary

Vals AI, Inc. is building the standard for evaluating misalignment in frontier models and agents. We seek an exceptional researcher to design and build the next generation of AI benchmarks that detect reward hacking, scheming, deception, and multi‑agent safety issues.

Your work will directly influence how foundations models are evaluated. You will lead benchmark design and infrastructure, collaborate with major labs and enterprises, publish findings, and stay at the forefront of alignment

Qualifications

• Hands-on safety evaluations, misalignment research, or related work.
• Strong background in ML/DL and model training fundamentals.
• Excellent Python coding and ability to build evaluation environments.
• Ability to communicate complex ideas to technical and non-technical audiences.

Responsibilities

• Design and develop high-impact benchmarks and environments for real-world misbehavior.
• Ensure benchmarks are valid, reliable, and meaningful through research.
• Collaborate with labs and enterprises to align with safety evaluation needs.
• Analyze model behavior across benchmarks and present results.
• Publish findings and contribute to the safety/evaluation community.
• Stay current with LLM capabilities and alignment methodologies.

Skills

Safety research

Deep ML/AI background

Strong engineering

Communication

Learning agility

Tools

Python

PyTorch/TensorFlow

AWS

Job description

Vals AI, Inc. is building the standard for evaluating misalignment in frontier models and agents. We seek an exceptional researcher to design and build the next generation of AI benchmarks that detect reward hacking, scheming, deception, and multi‑agent safety issues.

Your work will directly influence how foundations models are evaluated. You will lead benchmark design and infrastructure, collaborate with major labs and enterprises, publish findings, and stay at the forefront of alignment