Meta
Research Engineer, Safety Evaluation
$219K–$301K
Menlo Park, CA +1 · Full-time · Posted Sep 9, 2026
- Location
- Menlo Park, CA
New York, NY - Experience
- 3+ years of industry research or research-engineering experience in ML/AI.
- Education
- Bachelor's degree in Computer Science, Computer Engineering, or a relevant technical field, or equivalent practical experience.
About the job
Leads safety evaluation for Meta’s AI systems by designing and validating evaluations, building continuous evaluation infrastructure, and setting measurement standards that inform launch decisions.
What you'll do
- Set technical strategy and build evaluations for safety-critical behaviors across model families and modalities.
- Develop and operate scalable infrastructure, datasets, dashboards, and reporting for continuous safety evaluation.
- Translate policy and regulatory requirements into measurable criteria and communicate results to teams and leadership.
What they ask for
- 3+ years of industry research or research-engineering experience in ML/AI, including hands-on work with LLMs, multimodal models, or NLP.
- Experience designing and validating ML evaluations or benchmarks, including assessing metric reliability and failure modes.
- Experience building reliable production-grade or research infrastructure at scale.
- Programming experience in Python and hands-on experience with frameworks such as PyTorch.
Nice to have
- Experience evaluating agentic systems, including multi-step tool use, autonomy, and oversight.
- Peer-reviewed publications in evaluation, alignment, or AI safety; experience with distributed training, adversarial evaluation, or monitoring systems.
Pay
$219,000/year to $301,000/year (base salary)
Tools
Python, PyTorch
AI focus
Evaluates safety of frontier AI systems, including LLMs, multimodal and agentic systems; develops safety metrics, benchmarks, adversarial evaluations, and continuous evaluation infrastructure.
Model-written summary · the original posting is authoritative