Senior ML / Perception Engineer (Embedded Vision)

At TreeX Robotics, we are redefining the Drone. We are moving beyond contactless flight to interactive, autonomous flying robots that perform real physical work. Our first mission: Bringing our drones to handle the delicate, physical labor that wine grapes demand. As we scale, we are looking for a core team member for our AI development team. 

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 and guidelines  
• 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 basic export, validation, and debugging  
• Collaborate closely with software, robotics, and algorithm engineers to integrate perception into the system  

What We’re Looking For 

• 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)  

What Success Looks Like  

• A reliable perception model running on embedded hardware in real conditions  
• A data pipeline that continuously improves model performance over time  

Why This Role 

• Work on real-world problems, not benchmark datasets  
• Own a critical part of the system end-to-end  
• High autonomy and direct impact on product performance