Fully Funded DPhil Studentship in collaboration with APEX:E3 in Parameter-Efficient Fine-Tuning of Large Language Models

Page 1

Fully Funded DPhil Studentship in collaboration with APEX:E3 in Parameter-Efficient Fine-Tuning of Large Language Models Student fee status: Home Supervisor: Peter Minary in collaboration with APEX:E3 Start Date: October 2026 The Department of Computer Science at the University of Oxford is offering one fully funded DPhil studentship for home students, to commence next academic year (October 2026). The project will focus on parameter-efficient fine-tuning and adaptation of large language models, under the supervision of Peter Minary in collaboration with the APEX:E3 team (https://www.apexe3.com/). The successful applicant will join the AI for Computational Science team (https://www.cs.ox.ac.uk/people/peter.minary/), whose research interests include AI/ML, optimisation, probabilistic modelling, scientific computing, and language- model-based approaches for science and beyond. The project is particularly well suited to candidates interested in domain-agnostic AI/ML method development, efficient adaptation of foundation models, and the reliable incorporation of specialist knowledge into LLMs. The successful candidate will have the opportunity to collaborate with other members of the team and of the department. The studentship is offered in collaboration with APEX:E3, whose work includes agentic AI infrastructure and ALICE, an AI platform for capital markets. From AI agents in finance to LLM-based systems for scientific applications such as CRISPR- GPT, foundation models are creating opportunities across very different domains. These domains raise shared methodological questions around parameter-efficient adaptation, reliability, evaluation, and the integration of specialist knowledge. This project will explore new strategies for parameter-efficient fine-tuning, domain adaptation, and evaluation of large language models. The collaboration combines APEX:E3’s computational platforms and expertise in agentic AI with the AI for Computational Science team’s focus on problem-driven AI/ML method development for complex scientific systems. The aim is to develop approaches that are not tied to a single application area, but that can transfer across domains including finance, molecular biology, genomics, and broader computational science. The successful candidate will have the opportunity to work at the interface of modern AI/ML, foundation models, scientific computing, and real-world applications. The project would suit an applicant with strong interests in large language models, machine learning methodology, efficient fine-tuning, representation learning, evaluation, and/or agentic AI systems. Experience with LLMs, deep learning frameworks, and strong programming skills would be highly desirable. The studentship will provide an annual stipend (at the UKRI rate) at least £21,805 per annum for 3.5 years (42 months). The studentship will also cover the costs of course tuition fees at home level. Research students in the Department of

Page 2

Computer Science are also provided with a fund for travel to conferences and workshops (subject to approval). Applicants must satisfy the usual requirements for studying for a doctorate at Oxford, and apply online by Wednesday 1 July 2026 via the DPhil in Computer Science, quoting studentship reference 26-CS-PM. Candidates are expected to have a strong computational and/or mathematical background, with undergraduate studies in Computer Science, Mathematics, Physics, Engineering or related disciplines. Candidates must also have good writing, communication and presentation skills (see the University's web pages on the DPhil in Computer Science for details). For further information or for informal discussions about suitability, please contact Peter Minary (peter.minary@cs.ox.ac.uk). For further information about the studentship or the application process, please e-mail Computer Science Graduate Admissions.

Research areas