(Internal Only) Senior Engineer - Computer Vision / Machine Learning

Location: UK (London) preferred, Hungary considered
Contract: Permanent full-time
Level: 4

About the Role

You'll own the CV and physical-modelling layer within DemTech's tracking pipeline - working on top of ML-provided detection models to produce trajectory and positional outputs (e.g. trajectory estimation, motion reconstruction, 2D-to-3D reconstruction problems).

This role exists because tracking data is only as useful as the reconstruction layer that sits between detection and the outputs people actually rely on - dashboards, officiating decisions, performance insight. You'll own that layer within a small, fast-moving sports technology product team, working with real-world data.

You'll also be expected to work within DemTech's AI-first ways of working - using AI-delegated and AI-augmented development practices as a normal part of how you build, not as a separate initiative layered on top.

Key Responsibilities

Own the day-to-day delivery of the 2D-to-3D reconstruction pipeline - converting raw detections into positional and trajectory outputs using physics-based modelling (trajectory estimation, motion reconstruction, projectile physics), operating with autonomy within agreed direction.
Work directly with the ML discipline team on model performance - proposing and prototyping improvements where applied CV work surfaces opportunities, rather than only consuming their output. Strong performers here are expected to shape R&D-adjacent proposals, not just execute them.
Hold a genuine voice in technical decisions on algorithm design and data pipeline structure - contribute to architectural milestones and offer insight on peers' work, including alternative solutions and design tradeoffs.
Own the accuracy and reliability of tracking outputs across variable, real-world deployment conditions.
Support optimisation and deployment of models onto embedded, resource-constrained hardware, using deployment techniques such as TensorRT, ONNX, quantisation, pruning, and bottleneck profiling.
Use AI-delegated and AI-augmented development practices as a standard part of the role.

Key Attributes & Skills

Strong applied computer vision experience, with solid grounding in mathematical and physical modelling - trajectory estimation, motion reconstruction, projectile physics, or comparable 2D-to-3D reconstruction problems.
C++ required; Python experience is a plus for prototyping and tooling.
Working knowledge of ML techniques, with genuine interest in contributing to model improvement conversations and proposing R&D-adjacent ideas - core training and validation sit elsewhere.
Practical experience with model deployment and optimisation tooling (e.g. TensorRT, ONNX, quantisation, pruning, bottleneck profiling) for embedded or resource-constrained environments.
Experience with camera-based data sources, tracking pipelines, or spatial/temporal data.
Comfortable with ambiguity - this is a build-phase product with evolving scope.
Strong communication skills - able to work with data platform, backend, and frontend engineers on data contracts and outputs, and to explain complex problems and solutions clearly to others.

What This Role Is Not

Not a primary ML research role - core model training and validation sit with the ML discipline team or associated ML engineers, though close collaboration and proposing improvements is expected.
Not a data engineering role - a separate role owns storage, transformation, and API exposure of the outputs produced here.
Not a people-management role by default - this is an individual contributor position with genuine technical ownership of the CV/reconstruction domain, day-to-day and under agreed direction rather than final sign-off authority.