Samuel Oswald

Geospatial Data Scientist with 6+ years at VITO (Belgium). I developed a bespoke model management system for the MAPEO platform, operationalizing deep learning models for disease detection and plant phenotyping using multispectral drone and satellite imagery. I also developed time-series models that integrate harvester datasets with remote sensing inputs to enhance agricultural decision-making in the context of Horizon Europe.

My work includes implementing platform-level improvements such as automated Ground Control Point (GCP) detection and enhanced QA workflows to reduce labor costs. I specialize in transforming raw sensor data into production-ready repositories and data products for research partners and commercial agricultural stakeholders.

Technical Expertise

Computer Vision: Object detection, semantic segmentation, and change detection specifically tuned for top-down aerial and satellite perspectives.

Geospatial Model Operations (MLOps): Designing systems for model versioning, automated validation, and performance tracking across diverse sensor types.

Multi-Sensor Data Fusion: Synchronizing and integrating heterogeneous data streams from optical and other sensors for high-throughput phenotyping.

Spatial Data Engineering: Building end-to-end analytical pipelines and high-performance spatial databases for multi-temporal datasets.

Stack

Languages & Database: Python, Spatial SQL, PostGIS, PostgreSQL

Geospatial Engines: QGIS, GDAL/OGR, Agisoft Metashape

Scientific Frameworks: PyTorch, Scikit-learn, Xarray, GeoPandas

Deployment & Versioning: Git, Conda, Docker