Machine Learning Engineer — Aerial Image Classification

Role overview

Develop computer vision models that automatically identify assets, classify defects, detect changes, and estimate quantities in aerial inspection imagery. You’ll take models from data preparation and training through evaluation and deployment, supporting inspection needs across infrastructure types and operating both in clusters and at the edge.

Responsibilities

- Train, evaluate, and ship detection, classification, and segmentation models for imagery of roofs, solar arrays, transmission equipment, pipelines, and flare stacks. - Curate datasets and design augmentations for aerial imagery challenges such as varying scale, viewing angle, and combined thermal and RGB inputs. - Maintain training infrastructure and deploy models to cluster and edge environments. - Build evaluation tools that track regressions by class, region, and sensor. - Develop Python tooling for annotation, training, and deployment workflows. - Work with the edge team to reduce model size and improve on-device inference through quantization and pruning.

Requirements

- At least four years building and shipping production computer vision models. - Practical PyTorch expertise and familiarity with current detection methods and tools. - Strong Python engineering skills, including writing maintainable training software. - Experience with detection or segmentation; overhead or industrial imagery experience is especially relevant.

Nice to have

- Experience optimizing models for deployment with TensorRT or ONNX Runtime, including quantization.