O
Computer Vision Engineer
Oliver Brandt
Full-time · Mid-level · Amsterdam
About the role
I'm the CTO and I'm writing this myself because I want to be specific about what we're building and who will actually thrive here. We make industrial inspection systems for PCB manufacturers — automated visual defect detection on production lines running at 1,200 boards per hour. The core challenge is not model accuracy on a benchmark dataset. Our validation sets are pristine. The challenge is latency: we have 3 seconds per board, including capture, preprocessing, inference, and reporting to the line controller. Our current system runs at 4.1 seconds on average and we need to get it under 2.8 without meaningful accuracy regression. That's the specific problem. There are no silver bullets. We've already tried the obvious things — quantisation, batching strategy, model architecture changes. What we need now is someone who knows computer vision deeply enough to have non-obvious ideas. Someone who thinks about throughput the way a hardware engineer does, not just the way a machine learning engineer does. The team is six people. You'd be the second CV specialist. We ship real hardware. The work is technical and the feedback loop is immediate.
Responsibilities
- Profile and optimise our current inference pipeline to reduce end-to-end latency from 4.1s to under 2.8s
- Evaluate and implement model optimisation techniques: quantisation, pruning, architecture search, kernel fusion
- Extend defect detection coverage to three new defect types being introduced in Q3 by a major client
- Improve our synthetic data generation pipeline for defect augmentation
- Document every meaningful optimisation decision — what you tried, what worked, what didn't, and why
Requirements
- 4+ years of computer vision engineering with production deployment experience
- PyTorch — you profile training runs, you read CUDA traces, you understand why a certain layer is your bottleneck
- TensorRT or ONNX Runtime for inference optimisation — practical experience, not theoretical familiarity
- OpenCV for image processing at the preprocessing and postprocessing stages
- Solid understanding of object detection and segmentation model architectures — you can explain trade-offs between choices, not just implement them
- Experience with latency-constrained production environments is a significant advantage
- Edge hardware deployment experience (Jetson, industrial GPU cards, or similar) is a bonus
Benefits
- Hard real-world engineering constraints — not a research problem, a manufacturing problem
- Full remote with optional access to our hardware lab in Eindhoven for testing (flights covered)
- $100,000 – $125,000 base salary
- $1,500 hardware and tooling budget
- Small team, fast feedback, no bureaucracy
Job Type
Full-time
Level
Mid-level
Language
English
Salary Range
$100,000 – $125,000
AI Expertise
AI & Machine Learning Engineers
Ready to apply for this role?
Create a free talent account in under 2 minutes.
- Apply to verified AI companies
- Get AI-matched job recommendations
- Message hiring managers directly
- Build your public AI talent profile