Research Fellow in Reliable AI Alignment

About us

As part of our work in Machine Learning, the UCL Electronic and Electrical Engineering invites applications for one (1) postdoctoral research position in Reliable AI Alignment associated with an EPSRC New Investigator Award Project on “Robust and Efficient Algorithms for Model-based RL”. We propose a comprehensive research project aimed at addressing critical challenges in reinforcement learning from human feedback (RLHF; in the context of large language models) to ensure the development of reliable and robust artificial intelligence (AI) systems.

The project will focus on proposing novel algorithms and methods, incorporating uncertainty estimates, expanding beyond binary reward feedback, and addressing different groups’ preferences. Additionally, we aim to develop both theoretical frameworks and software packages to support the practical implementation of these advancements.

About the role

The position entails the development of algorithms, theoretical frameworks, and code for aligning large language models through Reinforcement Learning from Human Feedback (RLHF). This role demands expertise in leveraging recent advancements in active learning, robust optimization, and uncertainty quantification techniques. The successful candidate will collaborate closely with a team consisting of PhD students and Research Assistants (RAs).

Main duties:

• To develop novel RLHF algorithms and methods with a focus on reliability, safety, and alignment (e.g., to diverse user groups).
• To lead the drafting and submitting of papers to machine learning and AI conferences and journals.
• Presenting the research at international conferences and project meetings.
• To contribute to the preparation and drafting of research bids and proposals.

The post is available from January 2025 for 18 months in the first instance. Further funding to support the post may be available. The salary available is Grade 7.30 - 33 (£42,099 - £45,521 per annum, inclusive of London Allowance).

About you

Applicants should have a PhD degree (or about to submit) in a relevant subject area, or similar experience demonstrated via publications in international journals / conferences or patents. Knowledge of Machine Learning, Reinforcement Learning, Large language models is required, as well as strong programming skills in Python/PyTorch/ JAX or other relevant languages. Experience of working with statistical tools and techniques is also essential.

Knowledge of Statistical Learning Theory and Decision Making under Uncertainty is desirable.

If the successful candidate has not yet been awarded their PhD, appointment will be made as a Research Assistant (Grade 6B). Regrade to Grade 7 will be actioned in receipt of their PhD award. Salary range £37,332 to £39,980.

Application details:

• To apply for the role, click the 'Apply' button
• Applications close on 19th September 2024 at 23:59
• A job description and person specification can be accessed at the application link, as well as contact details for enquiries about the post

What we offer

As well as the exciting opportunities this role presents, we also offer some great benefits.

Visit https://www.ucl.ac.uk/work-at-ucl/reward-and-benefits to find out more

Our commitment to Equality, Diversity and Inclusion

We particularly encourage applications from candidates who are likely to be underrepresented in UCL’s workforce. These include people from Black, Asian and ethnic minority backgrounds; disabled people; LGBTQI+ people; and for our Grade 9 and 10 roles, women.

Our department holds an Athena SWAN Bronze award, in recognition of our commitment to advancing gender equality.

Advert information