Senior Computer Vision Engineer, Merantix

• You are a highly skilled computer vision engineer with proven experience taking CV projects from research/prototyping into production. You're excited about solving hard technical problems in real-world industrial settings. You communicate clearly, work autonomously, and thrive in a fast-moving startup environment
• 5+ years of hands-on experience in computer vision, AI/ML or robotics projects
• Strong expertise in Python and/or C++ and modern ML frameworks (PyTorch, TensorFlow)
• Deep understanding of computer vision algorithms, e.g. ByteTrack, DeepSORT etc
• In-depth knowledge of CV tools and libraries: OpenCV, torchvision, YOLO, object detection/tracking frameworks
• Production experience: you've taken at least one CV project from concept through deployment to end users
• Advanced degree: MSc or PhD in Computer Science, Machine Learning, Robotics or adjacent
• Cloud experience: hands-on work with at least one major provider (AWS, GCP, Azure)
• Code quality: experience with code reviews, testing, and maintaining large codebases
• Fluent English (German not required)
• Experience with multi-camera systems, 3D tracking, or geospatial reasoning
• On-premise or edge deployment experience (inference at the edge, privacy constraints)
• Background in logistics, robotics, autonomy, or industrial automation
• Experience with real-time systems or performance optimization for latency-critical applications
• Hands-on data labeling & curation experience (understanding ground truth quality)
• Entrepreneurial mindset: you're comfortable with ambiguity, move fast, and bias toward action

Desirable

• As Senior Computer Vision Engineer, you'll be instrumental in taking our automation platform from early prototypes into production systems deployed across logistics networks. You'll report directly to our CTO/Co-founder and work alongside a growing engineering team
• This is a hands-on, high-ownership role for someone who wants to solve real-world problems at scale. You'll work on cutting-edge computer vision challenges real-time object tracking, multi-view 3D reconstruction, ID association, and behavioral anomaly detection in messy, industrial environments
• Multi-camera fusion - Calibrate and fuse overlapping feeds into a single spatial model of the yard in real time
• Tag-free re-identification - Associate the same asset across cameras using appearance and motion alone. No RFID, no beacons
• 3D scene understanding - Map 2D detections to real-world coordinates and maintain queryable spatial state over time
• Behavioral anomaly detection - Learned models that generalize across new sites, not brittle rule-based heuristics
• Training stack - Own the pipeline from raw site footage to deployed model: labeling, experiment tracking, model versioning, and automated validation before anything ships to a customer site
• Data flywheel - Design pipelines that turn live customer footage into training data at scale so model quality compounds with each new deployment
• Edge deployment - Optimize for on-premise inference on commodity hardware
• Production quality bar - Define, monitor, and own what "good enough" means for mission-critical CV
• Pilot site presence - Diagnose failures in the field. Operational realities feed directly into engineering decisions
• Technical leadership - Set standards for model development, code quality, and CV best practices as the team grows