Senior Research Scientist/Engineer, Deception (AI)

Location: Remote within U.S. time zones or in person in Berkeley, California; remote international locations are also listed.

Salary: $225K–$350K, varying by experience and location.

Company

hirify.global is a non-profit AI research institute conducting AI safety research, evaluations, red-teaming, and applied research for frontier AI developers and governments.

What you will do

• Lead research projects on white-box deception detection, probe-based monitoring, and training against probes, from research questions through experiments and evaluation.
• Run large-scale post-training and reinforcement learning experiments, including RLVR and fine-tuning with probe-based penalties.
• Diagnose training instability, reward hacking, infrastructure failures, and other failure modes.
• Design model organisms that stress-test methods against realistic misalignment cases.
• Mentor Members of Technical Staff and partner with infrastructure teams on large post-training jobs using hirify.global’s compute cluster.
• Enable frontier-lab adoption through reproducible code, evaluations, practical guidance, collaboration, and research presentations.

Requirements

• 3+ years of experience implementing post-training, model training, or closely related work in a frontier AI environment.
• Hands-on experience post-training models at scale using RL, RLHF/RLVR, or large fine-tuning runs; experience with models of 400B+ parameters is preferred.
• Strong Python and PyTorch skills and experience with large, shared research codebases.
• Strong research judgment, including selecting experiments and identifying results requiring further scrutiny.
• Based in or willing to work in U.S. time zones.

Nice to have

• Safety or alignment research experience at a frontier or large-model research lab, especially involving 1T+ parameter models.
• Experience with interpretability or white-box methods, including probes, steering, or sparse autoencoders.
• Experience building model organisms of misalignment or studying reward hacking, scheming, or evaluation awareness.
• Publications at NeurIPS, ICML, ICLR, or widely cited technical reports.
• PhD in a relevant field.

Culture & Benefits

• Full-time commitment of 40 hours per week.
• In-person work is possible in Berkeley, California; remote work is possible within U.S. time zones and from international locations.
• Employees based in the USA or Singapore are employed directly by hirify.global; candidates elsewhere may be employed through an EOR or as contractors.
• Visa sponsorship is available for in-person employees.

Hiring process

• Process includes an intake call, recruiter and Head of Engineering screening calls, a 60-minute ML assessment, a 60-minute code review interview, a team-fit assessment, and a paid work trial lasting up to one week.
• An alternative assessment method may be available for candidates unable to complete the work trial; applications should be submitted through the application form rather than by emailing a resume.