We are looking for a highly skilled ML/Perception Engineer to take ownership of our vision pipeline - from data to deployed models running on embedded systems.
This role is focused on building perception systems that actually work in the field, under real-world constraints.
What You'll Do:
Own the full perception pipeline: data → training → evaluation → deployment
Develop and improve detection and segmentation models using RGB + depth data
Build and maintain a data engine: define labeling strategies, drive dataset quality and consistency, handle noisy and imperfect data
Ensure models generalize to real-world conditions: changing lighting, occlusions, small / hard-to-detect objects
Deploy models to Jetson NX-based systems - ensure compatibility with TensorRT, perform export, validation, and debugging
Collaborate closely with software, robotics, and algorithm engineers to integrate perception into the system
Requirements: Strong experience building and deploying computer vision models (detection + segmentation)
Hands-on experience with training pipelines and dataset iteration
Proven ability to improve performance with imperfect / noisy data
Experience working with real-world vision problems (not only curated datasets)
Practical understanding of deployment constraints (latency, memory, model size)
Experience exporting models to ONNX / TensorRT (or similar)
Nice to Have
Experience with RGB + depth / 3D perception
Background in robotics or embedded systems
Familiarity with edge devices (Jetson family)
This position is open to all candidates.