Postdoc Researcher — AI Safety for Mental Health Interviews

University of Oxford; Oxford; Hybrid; GBP 36,000 - 44,000; Full time

Job summary

The University of Oxford's Department of Psychiatry at Warneford Hospital seeks a postdoctoral researcher to advance AI safety work for the ARIADNE consortium. This full-time, two-year position focuses on developing LLM supervisor modules to detect risky patient interactions and support clinical debriefing.

The role requires a PhD in a quantitative field and experience with Python, PyTorch, and evaluating AI systems, with flexible hybrid work and at least three days on site per week.

Qualifications

• PhD in a quantitative field (ML, CS, computational neuroscience, mathematics) or near completion.
• Proven experience developing and evaluating LLM-based systems using Python and LLM APIs.
• Experience designing empirical evaluations of ML systems, incl. simulation-based methods and AI safety evaluation.

Responsibilities

• Develop LLM supervisor modules to detect inadvertent risky patient responses.
• Evaluate ARIADNE safety approaches in controlled testbeds with simulated patients.
• Collaborate with clinicians, ethicists and lived-experience researchers across partner universities and Google DeepMind.
• Publish and disseminate findings to advance AI safety in clinical interviews.

Skills

PhD in ML

Python

LLM APIs

Empirical evaluation

PyTorch

AI safety

Education

PhD or DPhil in a quantitative field

Tools

PyTorch

Job description

The University of Oxford's Department of Psychiatry at Warneford Hospital seeks a postdoctoral researcher to advance AI safety work for the ARIADNE consortium. This full-time, two-year position focuses on developing LLM supervisor modules to detect risky patient interactions and support clinical debriefing.

The role requires a PhD in a quantitative field and experience with Python, PyTorch, and evaluating AI systems, with flexible hybrid work and at least three days on site per week.