Machine Learning Engineer, Detection and Tracking

Who we are Helsing develops artificial intelligence-enabled capabilities to protect and defend democracies. We build Altra, an AI-powered drone software platform, and HX-2, our autonomous drone. We are growing our US operations, cultivating an ambitious and committed team of mission-driven professionals to apply their skills to solve challenging problems.

You will own the detection and tracking models that power Helsing's products — training, tuning, and deploying models against US-specific datasets. This is an applied ML role: you won't be writing research papers, but you will be expected to have strong intuition for model performance, data quality, and the practical trade-offs involved in getting detection and tracking systems to work reliably in production. You will manage the full model lifecycle — from assessing and curating training data through annotation, training, evaluation, and deployment to edge platforms. The day-to-day • Training and fine-tuning detection models (YOLO, DETR, Faster R-CNN, and similar architectures) on mission-specific datasets • Implementing and improving multi-object tracking pipelines (SORT, DeepSORT, ByteTrack, or similar) • Evaluating model performance: analyzing metrics, diagnosing failure modes, and iterating on data and model improvements • Managing the data pipeline end-to-end: assessing raw data, coordinating annotation, curating datasets, and implementing augmentation strategies • Optimizing models for deployment on SWaP-constrained and embedded platforms (quantization, pruning, TensorRT, ONNX export) • Collaborating with systems engineers to integrate models into the broader Altra platform • Working across sensor modalities as needed, including electro-optical, infrared, and other imaging sources You should apply if you • Have 5+ years of experience in applied machine learning or computer vision • Have a Bachelor's degree in Computer Science, Electrical Engineering, or a related field; Master's or PhD strongly preferred • Have production experience training and deploying object detection models — not just research or academic projects • Are proficient in Python and PyTorch or a comparable deep learning framework • Have strong intuition for data quality; you can look at annotated datasets, training curves, and evaluation metrics and know what's wrong • Have experience with the full model training lifecycle: data curation, annotation management, training, evaluation, and deployment • Have experience optimizing models for deployment on SWaP-constrained and edge platforms (TensorRT, ONNX, quantization) • Understand multi-object tracking and have implemented or worked with tracking algorithms in practice • Can read and contextualize scientific papers in computer vision and apply findings to production systems • Are a U.S. citizen with an active security clearance or the ability to obtain one