The project
How can we build AI systems that understand entire cities, from satellite height to street level, with sub-metre precision? This PhD will develop next-generation foundation-model-driven urban perception systems that fuse: high-resolution satellite imagery, aerial data, street-level imagery (e.g., Mapillary / survey data), LiDAR and 3D spatial data.
The goal: create scalable, intelligent mapping tools that support smarter, greener infrastructure planning for one of the UK's largest telecom providers. You will work at the frontier of:
· Vision-Language Models (VLMs)
· Generative AI and multimodal foundation models
· Weak supervision and open-set recognition
· Multi-modal sensor fusion across viewpoints and times
· Robust AI under seasonal and environmental variation
· Sub-metre geospatial reasoning
This is an opportunity to help define how Generative AI transforms national infrastructure management.
Academic Environment
This studentship is based at the Centre for Vision, Speech and Signal Processing (CVSSP). CVSSP is:
· The largest UK research centre in its field
· Ranked 1st in the UK for Computer Vision research )
· Home to world-leading research in computer vision, AI, multimodal learning, and robotics
You will join a vibrant and tightly knit team of more than 10 other researchers working on a range of cutting-edge research, including Generative world models, Event-based vision, Multispectral perception, 3D Gaussian Splatting and Industrial AI deployment. The team provides extensive mentoring and peer-support, as well as ample opportunity for collaborative projects.
Industrial Collaboration
This project is delivered in close collaboration with British Telecom Research. BT relies on accurate environmental data to plan and maintain national digital infrastructure. Inaccurate mapping increases planning costs, project timelines and disruption to its operations and communities.
You will be co-supervised by an industrial expert in Network Optimisation and AI from BT Research, ensuring that your work retains industrial relevance and real-world impact. This will also support access to data and operational datasets, direct engagement with engineers and potential routes into post-PhD research roles.
Why apply?
This project gives you the chance to:
· Work with cutting-edge multimodal foundation models
· Develop scalable AI for real cities
· Publish high-impact research
· Collaborate with a major UK infrastructure provider
· Contribute to smarter, greener network deployment
· Be part of a large, supportive PhD cohort
· Shape the future of infrastructure AI
Funding
This project offers an extremely generous funding package thanks to the industrial partner's involvement. Funding is available to support up to the maximum duration of four years if needed. Additionally, an enhanced stipend (25% beyond the standard EPSRC PhD stipend) is available with a generous Research Training Support budget to fund travel to international conferences.
Candidate profile
We are looking for a highly motivated candidate with a strong background in computer vision/machine learning and experience with PyTorch or similar frameworks. You should have an interest in multimodal learning and foundation models, and a passion for developing real-world AI impact. Experience with geospatial data, LiDAR, or satellite imagery is welcome but not required.
Funding is available only at the "home student" rate of tuition fees. Only applicants who are eligible to pay home fees should apply. If you are uncertain, please check here ).
If you are excited about pushing the boundaries of Vision-Language AI for solving real-world problems and bettering people's lives, we would love to hear from you Please find the official advert here