The project aims to leverage Artificial Intelligence (AI) techniques for mathematical reasoning by developing a novel approach to proving refinement propositions using a combination of LLMs and state-of-the-art theorem provers. Developments in this area have the potential to advance software verification, with expected practical impact across a wide range of applications such as in the engineering of mobile and autonomous robots.
Requirements
- You will have a PhD in formal methods
- experience on modelling and proof of algebraic refinement propositions using CSP
- their mechanisation in FDR
- Experience with use of theorem proving using Isabelle
- usage of Large Language Models (LLMs), including fine-tuning, is highly desirable.
- You will have experience with an object-oriented and or functional programming language.
Responsibilities
- You will work on the synthesis of novel refinement propositions and proofs obtained via state-of-the-art theorem provers and model-checkers.
- There will be a need to develop novel synthesis algorithms that can scale.
- There will be a need to fine-tune foundational models using the produced datasets and to evaluate their performance on benchmark problems.
- Novel reasoning techniques integrating LLMs and state-of-the-art provers will be developed, and there will be a need to formally justify their soundness.
- You will liaise with academics in the project and participate in the writing of papers.
Other
- The position is funded for up to 18 months.
- You will have the opportunity to further extend and develop your knowledge of formal methods, software engineering, and artificial intelligence whilst within the role.
- The University strives to be diverse and inclusive – a place where we can ALL be ourselves.
- We particularly encourage applications from people who identify as Black, Asian or from a Minority Ethnic background, who are underrepresented at the University.
- We offer family friendly, flexible working arrangements, with forums and inclusive facilities to support our staff.
