Safe Agentic/LLM Reasoning via Formal Verification The University of Manchester in United Kingdom

This PhD project at The University of Manchester explores innovative synergies between natural language-based Human-AI interaction, nature-inspired optimization, social impacts of Human-AI cooperation, and advanced machine learning in engineering. Generative AI is transforming how engineers access domain-specific knowledge and interact with AI, leading to enhanced productivity and the ability to tackle more complex design problems.

Two PhD positions are available at the intersection of formal software verification and Large Language Model (LLM) safety. The research focuses on extending state-of-the-art logic-based automated reasoning tools, such as ESBMC (https://github.com/esbmc/esbmc), to address safety and reliability challenges in agentic reasoning systems. Candidates will investigate novel approaches using abstract interpretation, model checking, constraint programming, and fuzzing techniques to ensure the safety and reliability of LLM-powered agentic systems.

The project is based in the Department of Computer Science and offers a stimulating environment for research in Artificial Intelligence, Cyber Security, Machine Learning, and Software Engineering. The successful candidate will join a diverse and inclusive research community, with opportunities for flexible study arrangements, including part-time options.

Funding: This 3.5-year PhD studentship is open to both Home (UK) and overseas applicants. The position provides an annual tax-free stipend at the UKRI rate (£20,780 for 2025/26, subject to annual uplift), and tuition fees are fully covered. The stipend is expected to increase each year. The anticipated start date is April 2026.

Eligibility: Applicants should have, or expect to achieve, at least a 2.1 honours degree or a master’s (or international equivalent) in a relevant science or engineering discipline. An English language certificate is required if applicable. Applicants must submit a supporting statement outlining their motivation, CV, transcripts, and contact details for two referees with official university/work email addresses. Incomplete applications will not be considered.

Application Process: Apply online via the university portal (https://uom.link/pgr-apply-2425). Specify the project title and supervisor name. Contact the supervisor before applying to discuss your motivation and academic background. After submitting your application, upload all supporting documents as required.

For questions about the application process, contact the admissions team at [email protected].

The University of Manchester is committed to equality, diversity, and inclusion, actively encouraging applicants from all backgrounds and supporting flexible study arrangements.