AI Safety Researcher: Red-Teaming, Safe RLHF, Unlimited PTO

Bison; San Francisco (CA); On-site; USD 350,000 - 475,000; Full time

Benefits offered by this job

Health benefits

Dental benefits

Vision benefits

Unlimited PTO

Parental leave

Relocation support

Job summary

Thinking Machines in San Francisco is seeking a safety researcher to advance safe and trustworthy AI. You’ll work at the intersection of research and engineering, exploring how models learn to refuse harmful requests and how training shapes behavior, with experiments guiding future development.

You will contribute across the stack—from data filtering and synthetic data generation to evaluation, red-teaming, and policy-based reasoning.

Qualifications

• Bachelor’s degree or equivalent in CS, ML, physics, math, or related field.
• Hands-on AI safety research experience in at least one safety area (RLHF/RLAIF, alignment, red-teaming).

Responsibilities

• Contribute across the development stack from data curation to safety-focused fine-tuning and red-teaming.
• Design experiments to study model safety boundaries and evaluate training outcomes.
• Build data filtering pipelines and quality classifiers to influence model pre-training.
• Develop safety evaluations and measure model behavior on long-horizon tasks.
• Generate synthetic data for training and evaluation and surface failure modes.

Skills

AI safety research

Python programming

Communication skills

Education

Bachelor’s degree or equivalent in CS/ML/Physics/Math

PhD in related field (optional)

Tools

PyTorch

TensorFlow

JAX

Job description

Thinking Machines in San Francisco is seeking a safety researcher to advance safe and trustworthy AI. You’ll work at the intersection of research and engineering, exploring how models learn to refuse harmful requests and how training shapes behavior, with experiments guiding future development.

You will contribute across the stack—from data filtering and synthetic data generation to evaluation, red-teaming, and policy-based reasoning.