We’re hiring for a Raw Ventures portfolio company — a computer vision platform that processes millions of images from real-world deployments and turns them into actionable insights for industry partners. Series A.
The role
You own the CV detection stack end-to-end — models, data, labeling and validation processes, evaluation methodology, production serving. You close the loop with the validators team and stay close to customers and product to keep model work tied to real-world value. Senior autonomy: you set direction and ship.
Stack
PyTorch • YOLO 11 • Roboflow • Triton • CUDA • AWS • EKS • S3 • Docker • Claude Code
What you’ll do
- Own the full model lifecycle — data, labeling, training, evaluation, deployment, monitoring, feedback
- Improve detection/segmentation models across diverse real-world conditions, lighting, environments
- Build labeling, testing, and validation processes with the validators team — taxonomies, guidelines, QA loops, active learning
- Define evaluation methodologies; identify weak spots and fix them
- Stay close to customers and product — what they pay for shapes model priorities
- Productionize models on Triton — latency, throughput, cost
- Drive research direction: pretraining, architectures, multi-stage pipelines
What we expect
- 5+ years CV/ML with senior depth
- Deep understanding of the full CV model lifecycle
- Track record of high-quality production CV models — robust across real conditions, edge cases handled
- Methodological eye — you spot when evaluation, labeling, or training is broken and fix it
- Product and business sense — you understand how models make money, don’t chase accuracy that doesn’t move the business
- Hands-on YOLO (YOLO 11 ideal, any recent version counts)
- Experience designing labeling/annotation processes with annotation teams
- Roboflow workflows or comparable (dataset versioning, labeling, augmentation)
- Triton deployment and optimization (or comparable serving infra)
- MLOps fundamentals — experiment tracking, model versioning, evaluation, production monitoring
- Strong PyTorch, production-grade Python (not just notebooks)
- Russian — fluent or native (required)
- English B1+
- Central European working hours
Nice to have
Active learning, semi-supervised methods • Self-supervised pretraining for domain adaptation • Model quantization / pruning / ONNX / TensorRT • Scientific imaging or biology • Multi-camera / multi-view systems • Edge inference on devices
What we offer
- Fully remote, CET hours
- Real product impact at scale
- Direct contact with leadership and engineering team
- AI-augmented development culture
- Competitive compensation, discussed individually
Originally posted on