Research Fellow in AI-Assisted Formal Verification

The University of Sheffield is a remarkable place to work. Our people are at the heart of everything we do. Their diverse backgrounds, abilities and beliefs make Sheffield a world-class university. We offer a fantastic range of benefits including a highly competitive annual leave entitlement (with the ability to purchase more), a generous pensions scheme, flexible working opportunities, a commitment to your development and wellbeing, a wide range of retail discounts, and much more.

Find out more about our benefits (opens in a new window) and join us to become part of something special. Overview   We are seeking an ambitious researcher to join a major new project funded by the Advanced Research and Invention Agency (ARIA), combining formal verification, cybersecurity and AI. The post offers an unusual opportunity to work at the intersection of Isabelle/HOL, the seL4-verified microkernel, information-flow security, and AI-assisted theorem proving.

The project aims to develop a formally verified reference monitor on top of seL4 for the secure containment of AI agents. We will develop new mechanisms for dynamically controlling agents' capabilities and information flows, together with machine-checked security guarantees. In parallel, we will investigate how modern AI techniques can accelerate large-scale formal verification, developing AI proof agents that can maintain, extend and refactor the seL4 proof base in Isabelle/HOL.

You will join a highly collaborative international team spanning the Universities of Sheffield, Surrey and Melbourne, bringing together expertise in Isabelle/HOL, seL4, information-flow security, program logics, and neurosymbolic AI. The project is exceptionally well resourced, including substantial funding for access to state-of-the-art AI models and computing infrastructure. We particularly welcome applicants with strong expertise in Isabelle/HOL or other interactive theorem provers, formal verification and security, or neurosymbolic AI and AI-assisted reasoning.

Deep expertise in Isabelle/HOL will be especially valued, and we encourage outstanding Isabelle researchers to apply even if their career stage is less senior than might normally be expected for a Grade 8 research position. For informal enquiries about the project or the position, please contact Professor Andrei Popescu at [email protected]   Main duties and responsibilities   Conduct research of international standing in areas relevant to the project, including formal verification, information-flow security, interactive theorem proving, and AI-assisted reasoning.

Independently develop and pursue technically ambitious research ideas, determining research objectives and selecting appropriate methods and approaches within the overall aims of the project. Take primary responsibility for substantial technical research tasks according to the postholder's expertise, from their formulation and implementation through to evaluation and dissemination. Contribute to the development and formal verification of an agent-aware reference monitor on top of seL4, supporting dynamic control of the capabilities and information flows of AI agents.

Develop and mechanise security models, program logics and information-flow properties in Isabelle/HOL, and contribute to the modernisation, extension and maintenance of the existing seL4 proof infrastructure. Develop and evaluate AI techniques for large-scale interactive theorem proving, potentially including supervised fine-tuning and reinforcement learning using formal verification feedback, and AI agents for proof generation, repair and refactoring.

Contribute to the development of Isabelle infrastructure for AI-assisted theorem proving, including proof-data extraction, interaction with proof states, and integration of AI agents into Isabelle-based verification workflows. Work closely with researchers and research software engineers at Sheffield and with project partners at the Universities of Surrey and