Projects 2025

Explore the active research projects led by students and researchers in the Practice-Oriented AI CDT.

AI-enabled cognitive assessment development

A second student team worked with Cognisess to investigate how can cognitive assessments be developed faster and more efficiently through the application of AI based approaches. The team applied Item Response Theory based approaches to game-based psychometric testing, enabling simulation, adaptive testing, and data-driven item design. They were able to demonstrate the potential to reduce costly pilot trials while improving reliability, opening new possibilities for scalable, AI-enabled digital assessment platforms.

Researchers: Nyamvula Chakaya, Gonzalo Cancio, Sophie Maw, Noah Barker and Russel Summers

More Information →

Mapping economic activity through online company presence

A third team of Pro-AI students asked whether it was possible to map the UK economy directly from the company presence on the Web? Working with researchers from the School of Geography, the project linked Companies House records with real websites operated by UK businesses. Combining search-based discovery, data extraction, and large language models, the system resolves complex company–website matches at a national scale with high accuracy and coverage comparable to commercial databases at a fraction of the cost. The approach could unlock exciting new ways for researchers, policymakers, and businesses to analyse the evolving digital economy.

Researchers: Prya Kharbanda, Tom Heslin, Isobel Higgins and Sanjukta Bhattacharya

More Information →

AI-optimisation of video recommendation systems in online streaming services

Working with Amazon, a team of students explored how can streaming platforms recommend content that feels both relevant and fresh? This project investigated next-generation recommendation systems that were able to balance recommendation accuracy with discovery of new items. Using real movie data, the team combined advanced machine learning approaches to produce a ranking strategy that dynamically balances relevance and diversity, resulting in more engaging video recommendation “slates.” The work further demonstrated the exciting potential of prompt-based recommendations where viewers can simply describe the kind of things they want to watch highlighting new opportunities for industry to build more engaging, personalised content discovery experiences.

Researchers: Abby Morris, Ed Daniels, Guoda Laurinaviciute and Ikechuwu Ofodile

More Information →

Reinforcement Learning for Trustworthy and Explainable AI in Decision Support for the Aerospace domain

This project aims to develop a Trustworthy AI (TAI) demonstrator, specifically leveraging Explainable AI (XAI) techniques. The core challenge is to design, prototype, and evaluate an approach that is not only effective in performance, but also demonstrably and appropriately trustworthy.

More specifically, the objective is to explore opportunities for reinforcement learning within this space to address aerospace decision making problems. RL is a powerful paradigm in machine learning where an “agent” learns to make sequential decisions by interacting with an “environment.” The agent’s goal is to maximise a cumulative reward signal, discovering an optimal policy - a mapping from states to action - through trial-and-error. This approach is highly effective for complex decision-making problems, such as navigation, resource allocation, and control systems, which are common in the aerospace domain.

The challenge: Select a public domain simulation environment (e.g. Gridworld or similar AI Gym environment) that is representative of the decision-making problems in the aerospace domain. Then train Reinforcement Learning agents using the environment(s) and select existing algorithms as a benchmarking exercise. Develop a trustworthy AI / Explainable AI technique, aimed to build appropriate trust of the agent and its behaviour with users. Then evaluate the approach. Demonstrate how the technique could be applied to other cases (i.e. simulation environments, problems, etc).

Researchers: Gonzalo Cancio, Tom Heslin, Guoda Laurinaviciute and Noah Barker

More Information →

Facilitating trusted interaction with AI-enabled smart thermostat control for heat pumps

Passiv UK is a software company that produces the Passiv Smart Thermostat (PST). The PST is designed to be like any other smart thermostat (easy to use, with both hardware interfaces in the home and an App) but is the only such product in the UK specifically designed for heat pumps. Our software uses Bayesian machine learning to characterise the thermal properties of a home and uses this information to optimise control of the heat pump (Model Predictive Control), enabling efficiency improvements, effective use of time-of-use tariffs, and in the future demand response.

One of our missions is to accelerate the deployment of low carbon heating by making heat pumps easy to use and understand. But many aspects of heat pump operation can be counterintuitive and it can be hard to get users to trust our system and get the most out of it. For example, understanding why the heat pump is running at unexpected times (heat pumps are most efficient running constantly, unlike gas boilers), or how to set up their heating schedules (many users unnecessarily set overcomplicated settings to reflect things the system is already doing automatically).

Project Aim The aim of the project is to improve people how people are able to understand and interact with and trust our AI system. This will entail understanding how people currently interpret and interact with the system, and how we can improve interpretability of model outputs to create trust in them and better align user behaviour to intended system behaviours. There is rich scope to combine interpretable ML and data analytics with appropriate UX design that supports improved human-AI interaction.

Researchers: Abby Morris, Nyamvula Chakaya, Isobel Higgins and Prya Kharbanda

More Information →

Human-Centred GeoAI: automating national mapping updates through interactive change detection

This project focuses on operating systems research and implementation aspects carried out by the team members, exploring low-level components and architecture constraints.

Researchers: Ed Daniels, Russell Summers, Sophie Maw and Ikechukwu Ofodile

More Information →