Senior AI Safety Scientist - Deception & Monitoring Lead

Far Ai; Berkeley (CA); Hybrid; USD 150,000 - 210,000; Full time

Job summary

FAR.AI is seeking an experienced research scientist/engineer to join our deception research initiative in Berkeley and across U.S. time zones. You will lead research projects from proposal to implementation, run post-training model evaluations at scale, and mentor junior staff.

This role blends rigor, collaboration, and practical impact, with opportunities to influence frontier AI safety. You will design experiments, publish findings, and collaborate with safety teams and infrastructure to

Qualifications

• 3+ years of implementing post-training or closely related efforts in a frontier AI environment.
• Hands-on post-training models at scale (RL/RLHF or large fine-tuning runs).
• Strong Python and PyTorch, and experience in large, shared research codebases.
• Strong research judgement to know which experiments to run next and scrutinize results.
• Based in or willing to work in U.S. time zones; Berkeley in-person preferred.

Responsibilities

• Lead research projects on deception detection, probe-based monitoring, and training against probes.
• Run post-training RL experiments, diagnose training instability and reward hacking.
• Design model organisms to stress-test methods against realistic failures.
• Mentor Members of Technical Staff to strengthen research judgment and rigor.
• Collaborate with infrastructure to run large post-training jobs on FAR.AI compute cluster.
• Ensure reproducible code, clear evaluations, and practical guidance for safety partners.
• Write and present findings for papers and the AI safety community.

Skills

Post-training models

Python

PyTorch

Research judgment

Berkeley in-person

Education

PhD in ML/AI

Job description

FAR.AI is seeking an experienced research scientist/engineer to join our deception research initiative in Berkeley and across U.S. time zones. You will lead research projects from proposal to implementation, run post-training model evaluations at scale, and mentor junior staff.

This role blends rigor, collaboration, and practical impact, with opportunities to influence frontier AI safety. You will design experiments, publish findings, and collaborate with safety teams and infrastructure to