What You’ll Do 

• Design, implement, and train deep learning models for:  

• 2D/3D object detection 

• Semantic segmentation of imagery and point clouds 

• Scene understanding, structural analysis, or surface defect detection 

• Work with large datasets from sensors, cameras, drones, and lidar 

• Optimize models for real-time inference and edge deployment (Jetson, embedded platforms) 

• Build pipelines for data preprocessing, annotation, and training automation 

• Collaborate with embedded, back-end, and platform teams to integrate models into production 

• Evaluate performance using precision/recall, IoU, and other task-specific metrics 

• Stay current on SOTA architectures and research, and evaluate their real-world tradeoffs 

• Contribute to infrastructure for versioning, tracking, and automating model lifecycles (ML Ops) 

Technical Proficiency 

• Strong foundation in deep learning frameworks: PyTorch, TensorFlow, or JAX 

• Experience with computer vision libraries: OpenCV, Torchvision, Albumentations, etc. 

• Skilled in model optimization: quantization, pruning, or ONNX/TensorRT deployment 

• Familiar with 3D data formats and pipelines: point clouds, meshes, geospatial datasets 

• Strong Python skills; bonus for C++ or CUDA for performance tuning 

• Experience with experiment tracking (Weights & Biases, MLFlow, etc.) and GPU training workflows 

• Bonus: Familiarity with SLAM, multi-view geometry, or temporal models (video/time series) 

What We’re Looking For 

• Degree (MS or PhD preferred) in Computer Science, Engineering, or related field with focus on deep learning or computer vision 

• 3+ years of experience shipping models into production environments 

• Strong math and spatial reasoning — geometry, linear algebra, probability 

• Experience with noisy or unstructured real-world data — not just ideal lab datasets 

• Ability to work cross-functionally and iterate quickly in an applied R&D environment