Research Associate in Detecting Unverbalized Dangerous Capabilities

Join Imperial College London to investigate and defend against hidden dangerous capabilities in frontier AI models.

Best for:Best for postdoctoral researchers with a strong background in deep learning, PyTorch, and AI safety or large language models.

Imperial College London · London, United Kingdom · R2 · Recognised researcher

More professors at Imperial College London

AI safetyLarge language modelsChain-of-thought reasoningModel monitoringAdversarial machine learning

This is a postdoc position — not a PhD studentship.

Deadline

Nov 22, 2026

30 days

Funding

Salaried

£50.7k–£59.5k/yr

Start date

ASAP and before 8th April 2027

Full-time

Location

London

United Kingdom

Overview

Imperial College London seeks a Postdoctoral Research Associate to join the Safety & Alignment lab under the direction of Dr. Oana-Maria Camburu. The fully funded 12-month project investigates methods to detect and defend against emerging scenarios where large language models conceal dangerous capabilities within their chain-of-thought reasoning. The successful candidate will design novel algorithms, conduct collaborative research, publish in top-tier venues, and mentor students. Applicants are expected to hold or be close to completing a PhD in Computer Science, Mathematics, Physics, or a related field.

Postdoctoral Research AssociateAI AlignmentFrontier ModelsPyTorchDeep LearningSafetyImperial College London

What you'll do

Design and develop novel algorithms to detect and defend against unverbalized dangerous capabilities in frontier models.
Conduct independent and collaborative research projects within the Safety & Alignment lab.
Publish research outcomes in top-tier conferences and journals such as NeurIPS, ICML, ICLR, and ACL.
Implement research ideas and contribute to broader projects involving frontier models.
Participate in group activities, including weekly lab meetings and presentations.
Mentor and support PhD and MSc students in the research group.

Requirements

Degree required: PhD in Computer Science, Mathematics, Physics or a closely related discipline (or equivalent research/industrial experience)

Language: English; excellent written and verbal proficiency

Instruction: English

Publications: Publications in relevant and refereed journals and conferences such as NeurIPS, ICML, ICLR, ACL, EMNLP, TACL, or AIJ

Required skills

Practical research experience within an academic or industrial environment
Demonstrated research experience with frontier AI models
Proficiency in coding with deep learning libraries such as PyTorch
Familiarity with techniques including model checking, monitoring, and chain-of-thought reasoning
Track record of publications in relevant refereed journals or conferences

Nice to have

Experience with automated reasoning or verification techniquesPrior experience mentoring junior researchers or studentsActive engagement with the AI safety and alignment research community

Eligibility

Candidates who have not yet been officially awarded their PhD will be appointed as Research Assistant

Funding & terms

Type

Salaried

Amount

£50.7k–£59.5k/yr

Duration

12 months fixed-term

Programme

Imperial College London

How to apply

Deadline: 2026-11-22

Candidates must visit the Imperial College London jobs portal, search for vacancy number ENG04085, and submit a full CV with a publication list along with a research statement.

Documents required