Data Scientist, AI Safety & Risk Analytics

Meta; Menlo Park (CA); On-site; USD 180,000 - 230,000; Full time

Job summary

Meta seeks Data Scientists to join the Safety team within MSL (Meta Superintelligence Labs). You will build analytical foundations to advance personal superintelligence safely and securely, turning ambiguous risks into measurable systems across model evaluations and online monitoring.

Work with engineering, research, product, policy, and legal to design measurement and mitigation strategies. You will quantify trade-offs between user friction and safety risks, help operationalize harm prevalence,

Qualifications

• Bachelor's or equivalent in CS/CE or quantitative field.
• 6+ years analytics experience; 4+ years with a PhD.
• Proficient with SQL, Python, or R.
• Experience with frontier AI products or risks in Trust & Safety or Fraud/Risk domains.
• Ability to work in fast-paced, cross-functional environments.
• Strong background in quantitative/statistical methods.
• Master's or PhD in quantitative field.
• Commitment to responsible, ethical AI practices.
• Experience integrating AI tools to improve workflows.
• Ongoing AI skill development (prompt/context engineering).

Skills

SQL

Python

R

Threat modelling

Causal inference

Cross-functional collaboration

Ethical AI practices

Agent orchestration

Prompt/context engineering

Education

Bachelor's degree in CS/CE or quantitative field

Master's or Ph.D. in a quantitative field

Tools

SQL (querying)

Python

R

AI measurement tools

Job description

Meta seeks Data Scientists to join the Safety team within MSL (Meta Superintelligence Labs). You will build analytical foundations to advance personal superintelligence safely and securely, turning ambiguous risks into measurable systems across model evaluations and online monitoring.

Work with engineering, research, product, policy, and legal to design measurement and mitigation strategies. You will quantify trade-offs between user friction and safety risks, help operationalize harm prevalence,