Every CleverCam alarm is verified by AI before a vehicle rolls: a person or a shadow, a car or a cat, a threat or a resident. Those models run on small Linux hubs at the edge, with a second look in the cloud, on real-world footage at night, in rain, on cheap cameras.
You will own that pipeline end to end: data from the field, training and evaluation, optimisation for edge hardware, deployment and monitoring, and the next generation of models that go beyond detection.
What you'll do
- Own detection and classification models for people, vehicles and other objects on live camera streams.
- Build the data pipeline: collect, label and curate footage from the field, with the privacy controls that requires.
- Evaluate rigorously against real false-alarm and missed-detection cases, not only benchmark sets.
- Optimise models to run on low-power edge hardware: quantisation, pruning and efficient architectures.
- Deploy and monitor models in production and close the loop from field feedback to the next release.
- Explore what comes next: behaviour and scene understanding, multi-camera reasoning and event scoring.
What you bring
Essential
- Degree in Computer Science, Engineering, Mathematics or similar, with strong applied machine learning experience.
- Three or more years training and shipping computer vision models to production.
- Deep experience with a major deep learning framework and with object detection architectures.
- Strong Python and solid software engineering habits.
- Experience running models on constrained hardware, or a clear understanding of what it takes.
Advantageous, not essential
- Edge inference toolchains and NPU targets.
- Video analytics, tracking and re-identification.
- MLOps: experiment tracking, dataset versioning and model registries.
- Publications or open-source work.
Who you are
- Rigorous:
- trusts the confusion matrix over the demo.
- Practical:
- ships the model that works at 3am on a wet camera.
- Curious:
- reads the paper, then tests it on our footage.
- A communicator:
- can explain precision and recall to a control-room manager.
The opportunity
The role can grow into leading AI at CleverCam.