Research Scientist - Safety

TTN Talent; San Francisco (CA); On-site; USD 360,000 - 440,000; Full time

Benefits offered by this job

Equity

Job summary

TTN Talent in San Francisco is seeking a rare on-site researcher to lead a new safety evaluation direction for frontier models. This role centers on hands-on AI safety work, reward hacking and RL-based evaluations, and designing benchmarks to measure aligned and misaligned behavior.

You will collaborate with engineers to turn research into reproducible evaluations used by leading AI organizations. The position is on-site in San Francisco, five days a week, with a salary up to $400,000 plus

Qualifications

• Hands-on AI safety research for frontier models
• Experience with reinforcement learning or model evaluations
• Experience with designing benchmarks and evaluating misaligned behavior
• Experience with reward hacking, deception, sabotage, or AI R&D capability evaluation
• Ability to turn research into reproducible evaluation environments

Responsibilities

• Design benchmarks and evaluation environments for frontier models
• Collaborate with engineers to turn research into practical evaluations used by leading AI organizations
• Lead safety evaluation initiatives as part of an early-stage applied AI research company

Job description

• Do you have recent hands-on research experience in AI safety and alignment for frontier models?
• Have you made meaningful contributions to recognised benchmarks, published research or public evaluations?
• Do you have hands-on experience with reinforcement learning, post-training or frontier-model evaluations?
• Have you worked on areas such as reward hacking, deception, sabotage, self-improvement or AI R&D capability evaluation?
• Can you design rigorous benchmarks, identify weaknesses in evaluations and work with engineers to turn research into reproducible environments?

If you answer yes to the above, then this confidential role could be the one for you.

This is a rare opportunity to join an early-stage applied AI research company and establish a new safety evaluation direction focused on frontier models. The successful candidate will design benchmarks and environments that measure advanced model capabilities and misaligned behaviour, while working closely with engineers to turn research into evaluations used by leading AI organisations. The role is based on-site in San Francisco, 5 days per week. Salary on offer is up to $400,000 plus equity (potentially a little more for an ideal candidate)