AI Safety Engineer (Data)

Written by the T&S Network from ElevenLabs's posting.

Owns the datasets and evaluation workflows behind AI safety work at an AI voice company spanning its ElevenAgents, ElevenCreative and ElevenAPI platforms. The role turns real-world product data into training and evaluation datasets and builds processes to assess how models perform in production, working with the AI Safety and Policy teams.

What you'd do

• Own safety datasets end to end, from collection and cleaning through labelling, quality control, versioning and training readiness.
• Translate safety policy into consistent labelling and evaluation criteria with policy specialists.
• Design labelling processes, including sourcing, onboarding and overseeing external contributors to a quality bar.
• Build evaluation workflows that use real-world data to track model performance in production and surface issues.
• Develop lightweight Python and SQL pipelines and tooling, and agree with ML engineers what training-ready data means.

What it asks for

• Several years working as a data scientist or in a similar role with real-world datasets and ML models.
• Solid command of what makes a dataset good, from gathering and cleaning through sampling, labelling, quality checks and evaluation.
• Working proficiency in Python and SQL for analysis and data workflows.
• Strong critical thinking, comfortable with ambiguity and turning complex guidelines into scalable decisions.
• Experience in AI safety, trust and safety, content moderation, or fraud and abuse detection (preferred).
• Experience creating datasets or evaluation frameworks, or working on adversarial or safety-focused ML (preferred).
• Experience managing human-in-the-loop labelling, dataset versioning or ML pipeline tooling (preferred).

Our best summary of the posting, and it can be wrong. The employer's posting is what counts: read it in full on ElevenLabsnorth_east before you apply. If we got something wrong, report this listing.