Five new projects will explore how AI can help tackle problems ranging from technology-enabled domestic abuse and workplace safety to air pollution, plant conservation, and the reconstruction of medieval manuscripts.
The projects are the latest cohort of ai@cam’s AI-deas Sprint Programme, which supports interdisciplinary teams across the University of Cambridge to turn promising ideas into working prototypes. Over six months, each team will work with partners and prospective users to build and test an AI-enabled tool in a real-world setting.
This year’s projects span different fields, but share a practical focus: using AI to help people make better use of complex information. They include a tool to help domestic-abuse practitioners identify the digital routes through which perpetrators can reach survivors; a system for bringing air-quality evidence into local transport decisions; and an application that can help scholars reunite manuscript fragments scattered across collections around the world.
“AI-deas is designed to help researchers take a promising idea and work out what it would take to make it useful in the real world. This cohort brings together a range of expertise, but what the projects have in common is that they are being developed with the people who understand the problem area and are developing AI to serve a real need. Over the next six months, the teams will be able to build, test and learn quickly, and see where their ideas can go.”
The five projects are:
Supporting frontline responses to technology-enabled abuse
Digital technologies have created new routes through which perpetrators of intimate partner abuse can monitor, locate, and contact survivors. Frontline domestic-abuse practitioners increasingly need to understand these risks as part of safety planning, but many lack specialist technical knowledge.
Led by Hannah Pankow, this project will develop a practitioner-facing tool to help workers and survivors map their digital exposure and develop a staged safety plan. The system will bring together technical, legal, and safeguarding knowledge to support practitioners’ judgement, including identifying possible routes of access and situations where specialist help is needed. The tool will be co-designed and tested with Cambridge Women’s Aid and lived-experience advisers.
Second Ahead: Predictive Safety
In construction, manufacturing, and logistics, a momentary lapse in attention while operating heavy machinery can have serious consequences. Existing monitoring systems tend to identify visible signs of fatigue or respond once unsafe behaviour has already begun.
Tyler Holderness and team will develop Second Ahead, a contactless system that combines information about an operator’s physiological state with an understanding of what is happening around them. Using standard video feeds, the system will explore whether signals including gaze, pupil diameter, blink rate, and heart-rate variability can help identify emerging risks before an accident occurs.
The team will initially focus on forklift operations, testing the approach with industrial partners in manufacturing and construction. Raw video will be processed locally and discarded, with the project also examining how biometric monitoring can be designed around workplace privacy.
Mapping the quality of the air we breathe
Greater Cambridge is expected to grow substantially over the coming decades, creating difficult questions about how changes to transport and the built environment will affect air quality. Although the city already has a network of air-quality sensors, measurements are unevenly distributed and changes at a particular location can be difficult to attribute to a policy intervention rather than weather, traffic, or other factors.
Led by Dr William Lee in partnership with the Greater Cambridge Partnership, the project will build a street-level air-quality mapping prototype. It will combine atmospheric dispersion modelling with existing sensor measurements to estimate concentrations of nitrogen dioxide, alongside, importantly, the uncertainty around those estimates. The team will test the approach by examining the possible air-quality effects of extending 20mph speed limits on selected Cambridge roads.
The longer-term ambition is to give local decision-makers a way to compare proposed interventions and understand the strength of the evidence behind their likely effects.
A collaborative platform to accelerate manuscript reconstruction
Hundreds of thousands of fragments from ancient and medieval manuscripts are dispersed between libraries and collections around the world. Finding fragments that once belonged to the same book can involve painstaking comparisons by expert researchers.
Vince Beiler, John Dudley and team will develop an application designed to make that process faster. A model will analyse manuscript images and suggest likely matches, which scholars can then inspect and confirm. Their decisions will in turn be used to improve subsequent recommendations.
The project will initially work with the Cairo Genizah, a collection of around 350,000 medieval Hebrew and Judeo-Arabic manuscript fragments, around 60% of which are held by Cambridge University Library. The team will scale an existing prototype to work across the collection and create a shared workspace through which scholars can build and document reconstructed manuscripts.
BGsmartR Curator: helping botanic gardens make better use of their collections
The world’s botanic gardens collectively hold more than 100,000 plant species, including many threatened with extinction. They also hold large amounts of data about those living collections, but smaller gardens in particular often lack the specialist staff and tools needed to turn those records into decisions about what to conserve, acquire, or share.
Led by Professor Sam Brockington at Cambridge University Botanic Garden, BGsmartR Curator will build on an existing open-source analysis tool by adding a conversational interface. Curators will be able to ask questions about their collections in everyday language—for example, which threatened plants are most vulnerable to being lost from a collection, or which species should be prioritised for conservation.
The project will work with botanic gardens across four continents and with Botanic Gardens Conservation International, with the aim of making sophisticated collections analysis accessible to gardens without dedicated data expertise.
From prototype to practice
The AI-deas Sprint Programme was established by ai@cam to help Cambridge researchers test ideas that sit between academic research and real-world use. Teams receive funding alongside technical support from ai@cam’s machine-learning engineers, with an emphasis on building, testing, and learning with the people who might use the technology.
Over the next six months, the new cohort will develop their prototypes with partners ranging from local government and domestic-abuse services to construction companies, libraries, and botanic gardens.
Professor Neil Lawrence, Chair of ai@cam, said: “Universities have an important role in working out not simply what AI can do, but how new capabilities can be made useful in practice. These projects bring together technical research, deep domain expertise, and people confronting the problem directly. That combination is essential if we want AI to deliver meaningful public value.”
The teams will present what they have built—and what they have learned—at the AI-deas Sprint showcase in 2027.
The AI-deas Sprint Programme is part of ai@cam, the University of Cambridge’s mission to drive AI innovation that benefits science, citizens and society.