Data Scientist, Machine Learning

This is a fully remote position, open to applicants in Mexico.

📋 Description

• Construct and calibrate change-detection and anomaly detection models utilizing multi-temporal Sentinel-1/2 imagery across pipeline corridors.

• Establish site-specific "normal terrain" baselines and validate detections against a ground-truth event log, focusing on detection rate, lead time, false-positive rate, and AUC.

• Refine geospatial foundation models (such as Prithvi-EO or similar) using LoRA/PEFT techniques on a limited set of labeled data.

• Apply SAR methodologies for displacement analysis, including amplitude change, coherence, and pixel-offset tracking to assess movement of pipes and dunes.

• Create tracking systems for dune migration, utilizing optical flow and feature tracking to determine migration direction and mobility indices.

• Develop a robust data ingestion pipeline from Copernicus (CDSE/Sentinel Hub/STAC) and integrate optical, SAR, DEM, and ERA5 wind data.

• Formulate a labeling strategy that includes encroachment masks and severity levels, as well as a train/validation split that prevents data leakage.

• Transparently communicate findings and limitations to both technical and business stakeholders.

⛳️ Requirements

• Over 7 years of experience in applied data science and machine learning, particularly with hands-on geospatial remote sensing.

• Proficient in Python, including libraries such as numpy, rasterio/GDAL, xarray, scikit-image, and geopandas/shapely.

• Familiarity with optical and SAR data, including understanding of spectral indices, backscatter/dB, resolution trade-offs, and revisit times.

• Experience with deep learning frameworks like PyTorch and a track record of fine-tuning models through transfer learning and LoRA/PEFT.

• Knowledge of model validation and calibration methods, including ROC/AUC, thresholding, cross-validation, and techniques for managing weak or limited labels.

• Expertise in time-series and change-detection methods, as well as coordinate reference systems (UTM and reprojection).

🏝️ Benefits

• Health insurance

• Opportunities for professional development