Projects 2024
AI for discovery of biodegradable polymers for Ration Packaging
In an era where sustainable solutions are paramount, our first group project tackled the challenge of developing environmentally friendly materials for military applications. Traditional methods of polymer discovery are often time-consuming and costly so our team used machine learning to accelerate identification of biodegradable polymers suitable for military packaging. With preliminary findings promising, further refinement and industry collaboration could unlock significant benefits in the production of next-generation materials.
Researchers: James Brock, David Fox, Moise Guran and Berenika Ewart-James
Multi Agent Reinforcement Learning for Cyber Security Agents
With the growth of IoT and mobile devices, dynamic network topologies present new challenges for automated cyber defence systems. Our second project focused on developing cyber security agents for identifying cyber threats in environments where network conditions are in constant flux. Through this project, our doctoral students are laying the groundwork for more resilient and scalable AI systems capable of defending against emerging cyber threats in an ever-changing digital world.
Researchers: Jack Woodcock, Jake Chastney and Jono Vincent
Supporting Post-Production of Volumetric Performance Video with AI
A team of Pro-AI students, in collaboration with immersive tech company Condense, has developed a pioneering machine learning pipeline to help speed up post-production of volumetric dance performances. Using skeletal pose estimation and Temporal Action Segmentation (TAS) the team have developed models that can segment and classify different dance moves in complex breakdancing sequences. Why it matters? Editing 3D video performances is currently labour intensive, requiring manual tagging and camera coordination. This project shows how AI can streamline this process, enabling semi-automated editing with further potential to support real-time highlight generation, audience-personalised streams, automated camera switching and even live-triggered visual effects - not just for dance, but for concerts, sports, and live entertainment.
Researchers: Jack Woodcock, David Fox and Berenika Ewart-James
Unlocking Scalable IT Support with Agentic AI
A student team, in collaboration with global technology company Unisys, developed a cutting-edge agentic AI prototype for automating first-line IT support, tackling a major time-consuming challenge of enterprise service desks: ticket triage and response. Their proof-of-concept system used a multi-agent architecture powered by large language models (LLMs) and LangGraph to simulate key IT support workflows from categorising issues to suggesting solutions. This proof-of-concept highlights the transformative potential of agentic AI in IT Service Management: reducing costs, improving response times, and enabling human agents to focus on complex problems. The project sets a strong foundation for next-generation service desk solutions.
Researchers: Bezawit Lake Tilaye, James Brock, Jake Chastney, Moise Guran