About this project
Generic AI tools offer little value in specialist disciplines as they lack the domain knowledge and pedagogical awareness that meaningful learning support demands. Geo-LLM addresses this gap by developing a purpose-built AI assistant for geotechnical and geoenvironmental engineering education.
Fine-tuned on curated course materials from four UK universities and grounded through a structured knowledge base, the system provides formative feedback that surfaces common misconceptions, clarifies method assumptions, and directs students to authoritative sources.
Unlike commercial AI tools, every response is cited and aligned with intended learning outcomes. Multi-site student trials across Cardiff, Manchester, Surrey, and Glasgow will evaluate its impact on learning effectiveness and confidence.
The project delivers a replicable, openly accessible framework, demonstrating that institutions can harness AI as a quality-assured enhancement tool rather than viewing it as a threat to academic standards.
Objectives
The project will develop and deploy a domain-aware AI assistant fine-tuned for geotechnical engineering, coupled with a structured knowledge base to ensure accurate, cited responses. Its pedagogical effectiveness will be evaluated through controlled student trials across four institutions, measuring feedback quality, misconception identification, and learner preparedness. A key legacy output is a practical "LLMs for Education" guide enabling other disciplines to adopt the same approach.
Methodology
Research interns will collaborate in parallel: one curating disciplinary content and structuring the knowledge base, and the other building the technical infrastructure including a secure web interface and intelligent retrieval system. Industry partners validate the system against professional standards and regulatory requirements. Student trials with anonymised surveys provide empirical evidence of impact across England, Wales, and Scotland.
Contributors
The project is led by Dr Evan John Ricketts (Cardiff University) with academic collaborators Dr Fei Jin (Cardiff), Dr Xiaomin Xu (Manchester), Dr Benyi Cao (Surrey), and Dr Zhiwei Gao (Glasgow). Cardiff's Geoenvironmental Research Centre and Advanced Research Computing team (ARCCA) provide additional expertise. Industry partners John F Hunt Regeneration Ltd, Arup, and CIRIA contribute regulatory validation and real-world case studies.
Project lead:
- Dr Evan John Ricketts, Cardiff University
Partner institutions and project contributors
- Dr Benyi Cao, University of Surrey
- Prof. Peter Cleall, Cardiff University
- Dr Mark Einon, Cardiff University
- Dr Adam Fisher, John F Hunt Regeneration Ltd
- Dr Zhiwei Gao, University of Glasgow
- Dr Michael Harbottle, Cardiff University
- Dr Fei Jin, Cardiff University
- Ben Kidd, ARUP, Joanne Kwan, CIRIA
- Dr Arif Mohammad, Cardiff University
- Dr Javier Munoz Criollo, Cardiff University
- Dr Nia Owen, Cardiff University
- Dr Richard Sandford, Cardiff University
- Prof. Devin Sapsford, Cardiff University
- Prof. Snehasis Tripathy, Cardiff University
- Dr Xiaomin Xu, University of Manchester
Pressing Escape will cancel and close this dialog
Dr Evan Ricketts
Evan obtained his MMath degree in Mathematics at Cardiff University, before completing a PhD in Civil Engineering at the same institution. Shortly after completion, he began as a Lecturer where he leads modules on finite element theory and practice. His current research centres on developing numerical methods for modelling heterogeneous materials, with particular emphasis on porous media. This work integrates finite element methods, random field theory, plurigaussian simulation, and scientific machine learning to capture material variability and enhance predictions of material behaviour under varying stimuli. Evan has published on topics ranging from stochastic unsaturated flow in soils to machine learning applications in self-healing cementitious materials. He is passionate about advancing computational approaches that bridge mathematical rigour with practical engineering applications, a commitment reflected equally in his teaching, and welcomes interdisciplinary collaboration.