Head of AI

Dex; Greater London; On-site; GBP 90,000 - 130,000; Full time

Job summary

Dex partners with a group building a surgical intelligence platform that fuses computer vision, AI, and augmented reality to deliver real-time 3D navigation for neurosurgery.

This senior AI/ML architect will lead core model development, data pipelines, and federated learning, mentoring 2–3 engineers, with a London-based in-person work requirement.

Qualifications

• Shipped production AI/ML models with CV/segmentation or vision-language experience.
• Experience with medical imaging (DICOM, MONAI) or the ability to ramp quickly.
• Led 2-3 engineers and seeks to grow a full AI pillar.

Responsibilities

• Deliver a working prototype of the anatomy recognition model for neurosurgery by Q2 2028, ≥85% segmentation accuracy.
• Lead the surgical foundation model program from computer vision to vision-language-action models for device automation.
• Architect the data pipeline for kinematics and tissue-interaction data across all endpoints.
• Own the federated learning stack, deploying NVIDIA FLARE for privacy-preserving model improvement.
• Define the boundary and partner with the simulation team on Isaac Sim, synthetic data, and sim-to-real transfer.

Skills

Production AI/ML

Computer vision

Segmentation

Vision-language

Federated learning

Team leadership

London in-person

Tools

NVIDIA AGX

NVIDIA DGX

MLOps infrastructure

NVIDIA FLARE

Job description

This role is with one of Dex's trusted partner companies. We work closely with their teams to truly understand their culture, goals, and what they're looking for, so we can match you with the right opportunity and give you context about the role before you commit to a process.

The role

Forget incremental improvements. This company is building the surgical intelligence layer, a platform that combines computer vision, AI, and augmented reality to give surgeons and robots real-time 3D navigation. Their core differentiator: a surgical foundation model, trained on kinematics, tissue interactions, and surgeon intent captured across deployed endpoints. This isn't academic research; it's about shipping models that fundamentally change surgical outcomes.

The work

• Deliver a working prototype of the anatomy recognition model for neurosurgery by Q2 2028, hitting ≥85% segmentation accuracy.
• Lead the surgical foundation model program: from computer vision and segmentation, to vision-language, to vision-language-action models for device automation.
• Architect the data pipeline for kinematics and tissue-interaction data across all deployed endpoints.
• Own the federated learning stack, deploying NVIDIA FLARE for privacy-preserving model improvement.
• Define the boundary and partner with the simulation team on Isaac Sim, synthetic data, and sim-to-real transfer.

What You Bring

• You've shipped production AI/ML models. Computer vision, segmentation, or vision-language experience is ideal.
• Experience with medical imaging (DICOM, segmentation networks, MONAI) or a rapid ability to ramp into it.
• You've worked with real-time inference constraints: sub-20ms latency, deployment to edge hardware like NVIDIA AGX.
• You own DGX/RTX-class training infrastructure and MLOps. Federated learning experience is a strong plus.
• You've led at least 2-3 engineers and want to grow a full AI pillar.
• Based in or around London, or able to work in person 3+ days a week.