Software Architect - Tracking

Start free →

Or apply yourself on the official page →

About the role

This is the senior computer vision engineering authority for Hawk-Eye's tracking technology, and it is an individual contributor role. Our tracking sits underneath almost everything Hawk-Eye sells. Goal-line technology, semi-automated offside, electronic line calling, automated ball-strike. Live and post-game data. It runs across football, baseball, tennis, American football, ice hockey, cricket, basketball and rugby, on-premise, in the cloud and increasingly on edge devices, delivered by over 100 engineers in the Tracking and Automated Officiating value stream. You would be the person who decides how that is built. The role carries two hats. As Software Architect you own the software-level architectural design across tracking, auto-officiating and data delivery. As computer vision authority you lead the core domains: camera calibration, ball tracking, skeletal tracking, and the R&D that feeds them. It is a hands-on job. You set architecture for new tracking problems as they arrive, you safeguard the shared frameworks, and when a live delivery is at risk you are the escalation point who gets into the code. What you would own Architecture across the value stream. You are the lead architect for every tracking and auto-officiating product in T&AO. You set the initial architecture when a new tracking problem enters the business, run the Software Domain Owner model so shared libraries like calibration have named codeowners, and keep designs consistent across sports. You represent tracking in the Engineering Architecture Steering group and feed into architecture conversations across Hawk-Eye and the wider Sony Sports organisation. Part of that is judgement about reuse. Keeping value streams on a common tracking codebase pays off until it doesn't, and knowing where that line sits is the job. Computer vision direction. You are the subject matter authority. You decide where our vision processing and ML integration goes next, research new tracking technologies and bring them in, and build prototypes when a proposal needs proving rather than describing. Modern tooling, properly evaluated. We want agentic coding frameworks, LLM-driven development tools and automated code generation working inside heavy C++ and Python codebases, with guidelines that hold up. You would lead that adoption and decide what earns its place. Growing other engineers. You coach the Squad Tech Leads, guide junior architects, and run CV Algorithm talks, HawkTalks and tutorials. You also help us hire well, including technical assessment for CV and ML roles. The awkward bits. Operational health of ML training and synthetic data infrastructure sits with you, including escalating the gaps in ownership and support with options attached. What we are looking for The core of it is deep multi-camera computer vision. Camera calibration including intrinsics, extrinsics, distortion models, anti-wobble and bundle adjustment. Triangulation. 3D ball and pose estimation, skeletal tracking, continuous-time modelling, inverse kinematics. On top of that: • Expert cross-platform C++ across Windows and Linux, x86 and ARM, with CMake and real performance profiling ability • Strong working knowledge of GPU inference (TensorRT, Triton) and ML pipelines (PyTorch, ONNX) • Proven experience running tracking at scale in AWS, on-premise and on edge devices • Hands-on use of AI coding agents and LLM orchestration to accelerate algorithm R&D and refactoring in large codebases • A working understanding of generative AI, synthetic data generation and where they genuinely help real-time CV The parts that are harder to put in a bullet point. You can set architecture across many squads and sports and make pragmatic calls on reuse, cost and time to market. You write clearly: design documents, pro

View the official posting →