Senior AI Research Scientist

• Salary: competitive, with significant stock options for early team members, and a workplace pension scheme.
• Flexible working: we are open to hybrid arrangements for the right person.
• Annual learning budget: for conferences, training or development.
• Time off: 25 days annual leave plus public holidays.
• Parental leave: three months paid maternity, one month paid paternity leave.
• Impact and recognition: work on life-changing technology, showcased globally in major medical breakthroughs.
• All the tools you need: new equipment and tools to help you do your best work, from laptops to specialised XR devices and software.

We will be growing our package of benefits over time.

• Own a research stream. Take one of the streams below from open problem to validated, real-time method. You set the approach, the baselines and the metrics.
• Build and benchmark. Develop methods on our registered surgical video corpus and public datasets; define the benchmark suite that engineering uses to accept a method.
• Take methods to real time. Profile and optimise for edge GPU inference, and meet the handover criteria (accuracy, latency, robustness, documentation).
• Design and run evaluations. Bench, phantom, cadaveric and supervised clinical evaluations with pre-specified endpoints, alongside Clinical Affairs.
• Publish. First-author work at the leading venues in computer-assisted surgery and computer vision.
• Shape the data. Contribute to annotation protocols, curation and quality control for the multimodal operating-room data programme.
• Mentor. Supervise placement students and co-supervise sponsored studentships as the team grows.

• Real-time stereo surface reconstruction: low-latency depth and surface estimation from exoscope and microscope video, robust to the optics and clutter of live surgery.
• Foundation-model adaptation for surgical video: parameter-efficient adaptation of vision foundation models to depth, pose and scene understanding that generalises across scopes, sites and procedures.
• Deformable registration and intraoperative model update: keeping preoperative anatomy registered to the live field as tissue moves and deforms, with quantified error.

Tell us which one is yours.

• You finish things. A method is done when it runs in real time on our hardware and beats the benchmark, not when the paper is accepted.
• You are honest with data. You report the failure cases and keep measured and modelled results apart.
• You care about the clinical problem. You want to see your method in an operating room, and you will spend time there to understand it.
• You work well with engineers. You write code they can read and you respect the handover.

We are at the intersection of AR, AI and surgical robotics, backed by top investors and working with leading hospitals in the UK and US. If you want to be part of a small, elite team building groundbreaking technology from the ground up, this is your opportunity.