Prepare, curate, and manage training datasets from remote sensing sources (Sentinel-2, Sentinel-1, Landsat, lidar, NAIP) and field plot inventories.
Evaluate model performance using standard remote sensing accuracy metrics and field-based validation data.
Contribute to experiment design, hyperparameter optimization, and ablation studies in coordination with the Technical Lead ML Engineer.
Pipeline & Data Engineering
Integrate trained ML models into Vibrant Planet’s automated geospatial data pipeline as containerized, orchestrated inference services.
Build and maintain STAC (SpatioTemporal Asset Catalog) infrastructure for data discovery, cataloging, and access control of ML model inputs and outputs.
Design and implement larger pipelines composed of many smaller DAGs (Airflow), ensuring idempotency, observability, and fault tolerance.
Maintain and improve data ingestion, preprocessing, and quality control workflows for satellite imagery and ancillary datasets.
Monitor pipeline health and model drift; implement alerting and automated retraining triggers as needed.
Develop model cards for summarization of modeling methods and performance.