Protection Scientist Engineer, Integrity

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

Protection Scientists in Integrity build and run systems of defense against harmful usage of OpenAI's products, covering harms such as child safety, nation-state misuse, and violence alongside frontier harms. The team builds harm protections for new product and model launches and responds to bad actors on the platforms.

• Child safety and grooming

What you'd do

• Set protection strategy and priorities for a harm area or launch, trading off risk against available capacity.
• Create and keep up production detection with SQL, Python, classifiers and language models.
• Investigate emerging harm and abuse patterns with domain experts, separating harmful activity from legitimate use.
• Coordinate cross-functional delivery with investigations, policy, operations, enforcement and data science partners.
• Gauge leftover harm and how well detection and interventions perform, then tune the system against actual cases.
• Define data, detection and review requirements for product launches and investigate critical escalations after launch.
• Build reusable tools and documentation, using language models and automation with human oversight.

What it asks for

• Substantial knowledge of at least one harm area gained in integrity, T&S, security or investigations work.
• Proficiency in SQL and Python, with experience writing reliable production code for detection and data processing.
• Quantitative analysis and machine learning skills for measuring risk and evaluating system effectiveness.
• Track record of planning and shipping multi-team technical work.
• Clear written and verbal communication of technical findings and trade-offs to varied audiences.
• Experience using language models and automation for analytical or operational work (preferred).
• Additional ML or software engineering experience (preferred).

Worth knowing

• Participation in an on-call rotation that may involve resolving urgent escalations outside normal working hours.
• Some investigations may involve sensitive sexual, violent or otherwise disturbing content.

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