Machine Learning Engineer, Safety

Harrison Clarke; San Francisco (CA); Hybrid; USD 190,000 - 275,000; Full time

Job summary

Harrison Clarke is seeking a Machine Learning Engineer focused on safety for a frontier AI lab in the Bay Area. This role centers on making advanced AI systems reliable, controllable, and aligned as capabilities grow, with hands-on work at the edge of current capabilities.

You will work on evaluation, red-teaming, safety post-training, and guardrails, with a bias toward shipping solutions end-to-end in a compact, senior team. Hybrid Bay Area work arrangement offered.

Qualifications

• Strong ML engineering skills with hands-on experience in LLMs or foundation models.
• Experience in safety-focused areas such as alignment, evals, or post-training methods.

Responsibilities

• Build evaluation and safety-oversight systems for advanced reasoning.
• Perform red-teaming and adversarial testing to drive model improvements.
• Develop safety-focused post-training, reward modelling, and guardrails.
• Identify and mitigate failure modes in multi-step reasoning under ambiguity.

Skills

ML engineering

LLMs/foundation models

safety/alignments

red-teaming/evals

Job description

Machine Learning Engineer, Safety | Stealth Mode Frontier AI Lab | Bay Area

I'm working with a well-funded, early-stage stealth AI lab building genuinely frontier systems - and they're hiring a Machine Learning Engineer focused on safety.

The mission: make advanced AI systems reliable, controllable, and aligned as their capabilities grow. This is hands-on, unsolved-problem work at the edge of what's possible.

What you'd work on

• Evaluation and oversight systems for advanced reasoning and agentic behaviour
• Red-teaming and adversarial testing - turning findings into real model and training improvements
• Safety-focused post-training, reward modelling, and guardrails
• Identifying and mitigating failure modes in complex, multi-step reasoning

You might be a fit if you have

• Strong ML engineering skills and hands-on experience with LLMs / foundation models
• Work in one or more of: post-training (SFT/RL/RLHF), evals, red-teaming, alignment, or safety infrastructure
• A bias toward shipping and owning problems end-to-end in an ambiguous environment
• Real interest in the hard problems of frontier AI safety

Details

• Bay Area, hybrid
• Small, senior, talent-dense team - real ownership from day one