Zone 5 Technologies is redefining what's possible in unmanned aircraft systems, developing innovative autonomous solutions. They are seeking a Machine Learning Engineer to build production-grade computer vision pipelines and manage end-to-end ML data infrastructure for autonomous aerial systems, owning the full lifecycle from data curation through model deployment on edge hardware.
Responsibilities:
- Design, train, and optimize object detection and tracking models (YOLO, RT-DETR, ViT) for aerial platforms
- Deploy models on NVIDIA Jetson hardware using TensorRT, achieving real-time inference
- Integrate vision models into ROS2 autonomy stacks with efficient message passing and synchronized sensor data
- Profile and optimize model performance to meet compute, power, and latency constraints on embedded systems
- Build and maintain data pipelines for ingestion, labeling, versioning, and quality control of large-scale aerial imagery datasets
- Design active learning loops and data selection strategies to maximize model performance with minimal labeling effort
- Implement automated model evaluation frameworks and performance monitoring for deployed systems
- Create synthetic data generation pipelines and data augmentation strategies to improve model robustness
- Develop ROS2 nodes for real-time sensor processing, model inference, and downstream autonomy integration
- Work with GStreamer and Jetson multimedia APIs for efficient camera pipeline management
- Collaborate with perception and autonomy teams to ensure vision outputs meet system requirements
Requirements:
- Bachelor's in Computer Science, Electrical/Computer Engineering, Robotics, or related field (Master's preferred) – equivalent industry experience also welcome
- 4-6+ years of experience in machine learning with focus on computer vision applications
- Strong proficiency in PyTorch and modern deep learning frameworks
- Hands-on experience deploying models on NVIDIA Jetson platforms (Orin, Xavier, or similar)
- Solid understanding of model optimization techniques: quantization, pruning, TensorRT, ONNX
- Experience with ROS2 for robotics applications, including sensor integration and message handling
- Proven track record building and managing ML data pipelines and infrastructure
- Strong Python skills with experience in data processing libraries (NumPy, Pandas, OpenCV)
- Familiarity with MLOps tools and practices (versioning, experiment tracking, CI/CD for models)
- Experience with C++ for performance-critical vision processing
- Knowledge of CUDA programming or GPU optimization
- Familiarity with data labeling tools and annotation management systems
- Experience with synthetic data generation or domain adaptation techniques
- Understanding of aerial imagery characteristics and challenges
- Background in object tracking algorithms (Kalman filters, Hungarian assignment, multi-object tracking)
- Experience with GStreamer, Jetson BSP, or embedded Linux development
- Knowledge of model monitoring and performance debugging in production systems
- Familiarity with DVC, MLflow, Weights & Biases, or similar MLOps platforms