THE OPPORTUNITY
A role worth
a closer look.
Design automated assessments that look for deceptive or misaligned behaviour during frontier-model training. Run evaluation engagements with AI labs, investigate unusual results and improve the methods used between campaigns. Combine production Python engineering, quantitative analysis and clear reporting to turn model behaviour into actionable evidence.
Experience highlighted by the employer
- Strong Python software engineering and the ability to find meaningful patterns in messy evaluation data.
- Clear technical and non-technical communication; a formal degree or prior industry background is not required.
Location & eligibility
This listing covers San Francisco and London; the US opportunity is in person in San Francisco, with flexible working hours and some work-from-home arrangements. Visa sponsorship is considered in the US and UK but is not guaranteed.
Compensation
See the employer’s original listing for current compensation details. We do not estimate an undisclosed currency or salary.
Before you apply
This is TaskBull’s concise summary, not the employer’s full job description. Review the complete responsibilities, terms, accessibility information and privacy notice on the original application page.
Read the full employer listing ↗