AI Safety Practitioner – 100% Remote at Mercor

About the Role

Key Responsibilities

• Conduct systematic risk assessments, red‑team exercises, and adversarial testing on large‑scale language and vision models.
• Co‑create safety‑by‑design guidelines with product and engineering teams, ensuring seamless integration into agile development pipelines.
• Define, monitor, and report on safety metrics, delivering clear deliverables against tight milestones and regulatory timelines.
• Lead incident post‑mortems, communicate findings to stakeholders, and propose corrective actions using data‑driven storytelling.

Required Qualifications & Skills

• Minimum 3 years of hands‑on experience in AI safety, ML security, or related research, with a proven publication or project record.
• Proficiency in Python, TensorFlow/PyTorch, and familiarity with formal verification, interpretability tools, and prompt engineering.
• Excellent written and verbal communication, strong self‑management, and ability to thrive in a fully asynchronous environment.
• Dedicated home office meeting Mercor’s standards (high‑speed internet, ergonomic setup) and legal right to work remotely for global clients.

Why Join & Remote Benefits

• 100 % remote flexibility with core overlap windows for global sync, eliminating commute and supporting work‑life harmony.
• Competitive compensation up to $70 USD per hour, performance bonuses, and equity‑style incentives.
• Access to cutting‑edge AI research, continuous learning budget, and mentorship from industry pioneers.
• Health, dental, and vision coverage, plus a wellness stipend and paid time‑off policy tailored for remote teams.

Application & Interview Tips

• Tailor your résumé to highlight concrete safety projects, metrics achieved, and any open‑source contributions.
• Include a brief portfolio or GitHub links demonstrating red‑team scripts, safety dashboards, or research papers.
• Prepare to discuss a real‑world AI failure case, your diagnostic approach, and how you would mitigate it.
• Emphasize your remote‑work discipline: time‑zone coverage, communication tools mastery, and self‑directed learning habits.