About this role
Summary: Design, train, and deploy computer vision models for autonomous vehicle navigation and perception. Responsibilities include dataset creation, model optimization for edge devices, and flight testing integration.
As a machine learning engineer in the computer vision team you will be responsible for designing, training, evaluating and deploying machine learning models onto a large fleet of operational vehicles.
Minerva Defence is one of the fastest-growing defence technology companies in Europe. We create the software, hardware and networking that make unmanned systems effective on the modern battlefield, all ideated, designed and manufactured end-to-end in the UK and in active use today. We are deliberately low-profile: a team of around 80, small in headcount and outsized in effect. We look for people who are not just talented but genuinely committed to the work, the team, and the purpose behind it.
You’ll join the team responsible for the vision and autonomy stack on the vehicles. The work spans navigation, perception and guidance. Engineers will be expected to operate across the full range of the problem set from collecting/designing datasets, model selection/training, optimisation for inference on-edge, flight testing on real systems and monitoring/improving operational performance.
Responsibilities
- Build and improve the models behind vision-based navigation, perception and guidance.
- Design, collect and if necessary build annotation pipelines for real and simulated datasets.
- Design and train deep neural network based computer vision models
- Optimise models for deployment on edge compute
- Work with the flight control engineers so model outputs feed reliably into guidance
- Diagnose failure modes from flight data and fix them through data or model changes
Minimum:
- Hands-on experience training and evaluating deep neural networks on real-world image or video data
- Strong Python, plus experience with a major deep learning framework such as PyTorch
- Experience getting a model running on real hardware or in a production system
- Able to work from our London office on a hybrid basis
Preferred:
- Experience with visual navigation, visual odometry or SLAM
- Experience with object detection or classification
- Familiarity with guidance, navigation and control concepts
- Experience fusing data from multiple sensors or imaging types
- Experience deploying models to embedded hardware, including quantisation and/or distillation
- Eligible for UK SC clearance (UK citizen or 5+ years UK residency)
We offer competitive compensation including;
- Share options vesting over three years
- Bupa private medical insurance
- 6% employer pension contribution
- 25 days leave plus bank holidays
- Enhanced mat/pat leave
- Enterprise level AI tooling
- Relocation support/home office budget