Geospatial Data Scientist

Orcrist is building a next generation data intelligence platform using cutting-edge technologies. We're handling petabyte-scale data with sub-second queries. Our product is a Kubernetes‑based platform delivered as B2B SaaS or as a self‑hosted on‑prem solution, including air‑gapped deployments. We enable customers across defense, law enforcement, and enterprise to turn mission-critical data into actionable intelligence.

Role

Develop geospatial and remote-sensing methods that turn imagery and spatial data into reliable analytical products. You'll work in Python on change detection, object detection, segmentation, and spatial and temporal analysis. Your methods will span classical machine learning, statistical and image analysis, and deep learning architectures such as Vision Transformers (ViTs) and U-Nets, taking the approaches that prove useful from exploration through evaluation into repeatable platform capabilities.

The work includes open satellite time series and commercial optical, thermal, and SAR imagery. You'll help choose appropriate data and methods, establish what the results can support, and work with engineers to make them available to analysts inside Sentinel.

What you'll do

• Develop change-detection workflows, including Sentinel-2 time series, and distinguish meaningful change from seasonality, cloud and shadow effects, acquisition differences, and registration errors.

• Build and evaluate object-detection, segmentation, and classification methods for satellite imagery, using statistical techniques, classical image processing, and machine learning where appropriate. Evaluate state-of-the-art deep learning approaches pragmatically against simpler baselines, weighing accuracy, label requirements, generalization, inference cost, and real production viability.

• Assess the suitability of different sensors, resolutions, acquisition conditions, and processing levels for each use case. Extend methods across optical, multispectral, thermal, and SAR data as requirements develop.

• Design preprocessing and feature extraction with data engineers: quality masking, compositing, co-registration, normalization, spectral indices, and sensor-specific corrections.

• Combine raster outputs with vector and temporal data for spatial statistics, zonal analysis, anomaly detection, and comparison across areas and observation periods.

• Build or source reference datasets and evaluation protocols. Use spatially and temporally separated validation, measure false positives and missed detections, and examine performance across regions and sensors.

• Deliver traceable results with source references, timestamps, confidence or uncertainty measures, and documented limitations. Help analysts understand when a result needs closer review.