Geospatial Data Scientist at Orcrist Technologies

Job description

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. Package tested Python methods for repeatable batch processing or inference. Work with data and platform engineers on runtime, memory, monitoring, and integration into Sentinel workflows. About you Formal university training in geoinformatics, remote sensing, Earth observation, applied mathematics, computer science, or a related discipline, combined with substantial hands-on geospatial or remote-sensing experience. Strong Python and scientific computing skills, using tools such as NumPy, pandas, SciPy, xarray, GeoPandas, and Rasterio/GDAL. Practical machine-learning experience with scikit-learn and PyTorch or an equivalent framework, plus TorchGeo or related geospatial deep learning packages, including training, evaluation, and adapting existing models. A solid understanding of remote-sensing fundamentals: spatial, spectral, radiometric, and temporal resolution; coordinate systems; image alignment; and data-quality limitations. Experience with satellite image analysis and at least one relevant task such as change detection, segmentation, object detection, or land-cover classification. Sound statistical judgment around sampling, spatial autocorrelation, data leakage, class imbalance, uncertainty, and generalization to new places and acquisition conditions. An engineering-minded approach to research: versioned code and data, reproducible experiments, tests, and clear explanations of how a method performs and fails. Clear communication in English with both technical colleagues and domain specialists. Eligible to work in Germany. Nice‑to‑haves Experience working with thermal infrared imagery, processing and analysing SAR data, or combining observations from multiple sensors. Deep experience in one modality is valuable. A PhD in geoinformatics, remote sensing, Earth observation, or a related field. Experience with geospatial foundation models: fine-tuning models such as Prithvi or TerraMind, or using AlphaEarth Foundations embeddings for downstream analysis. Ability to evaluate whether these approaches improve on task-specific models with the available imagery and labels. Experience scaling geospatial analysis with Dask, Apache Spark/Sedona, or Zarr. Experience deploying and optimising GPU inference with PyTorch/CUDA, ONNX Runtime with TensorRT, or NVIDIA Triton Inference Server, including batching image tiles and managing GPU memory and throughput. Spatial SQL with DuckDB or PostGIS, STAC-based data discovery, and exploratory work in QGIS or geemap; producing COG and GeoParquet outputs for downstream GIS use. Experience working with commercial imagery from providers such as Airbus, Satellogic, SatVu, or ICEYE. Experience working in defence and intelligence environments or on related projects. Strong interest and practical ability in agentic software development: using coding agents to plan, implement, test, and review software, and keeping up with rapidly evolving tools, techniques, and trends. What we offer The opportunity to establish new analytical capabilities within an existing intelligence platform. A mix of scientific depth and practical delivery, with direct feedback from GEOINT specialists and analysts. Close collaboration with data engineers and software engineers to bring useful methods into production. Remote-first, Germany-wide: Work from wherever you do your best work, with regular team gatherings in Berlin and other off-site locations. Flexibility by default: Flexible working hours help you make work fit your life. Your setup, your way: Get a personal home-office equipment budget to create a workspace that works for you. 30 days of vacation: Take the time you need to recharge and come back with fresh energy. Keep growing: We invest in your personal and professional development. Get rewarded for impact: Performance bonuses are tied to agreed objectives and key results. A warm welcome: Every new team member gets a welcome goodie bag. Good people, good times: From summer and Christmas parties to regular team gatherings, we make time to celebrate together. A mission that matters: Work on challenges with tangible impact on public safety and national security.

Requirements

Must-haves

Develop change-detection workflows, including Sentinel-2 time series, and distinguish meaningful change from seasonality, cloud and shadow effects, acquisition differences, and registration errors.
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.
Package tested Python methods for repeatable batch processing or inference. Work with data and platform engineers on runtime, memory, monitoring, and integration into Sentinel workflows.
Formal university training in geoinformatics, remote sensing, Earth observation, applied mathematics, computer science, or a related discipline, combined with substantial hands-on geospatial or remote-sensing experience.
Strong Python and scientific computing skills, using tools such as NumPy, pandas, SciPy, xarray, GeoPandas, and Rasterio/GDAL.
Practical machine-learning experience with scikit-learn and PyTorch or an equivalent framework, plus TorchGeo or related geospatial deep learning packages, including training, evaluation, and adapting existing models.
A solid understanding of remote-sensing fundamentals: spatial, spectral, radiometric, and temporal resolution; coordinate systems; image alignment; and data-quality limitations.
Experience with satellite image analysis and at least one relevant task such as change detection, segmentation, object detection, or land-cover classification.
Sound statistical judgment around sampling, spatial autocorrelation, data leakage, class imbalance, uncertainty, and generalization to new places and acquisition conditions.
An engineering-minded approach to research: versioned code and data, reproducible experiments, tests, and clear explanations of how a method performs and fails.
Clear communication in English with both technical colleagues and domain specialists. Eligible to work in Germany.
Nice‑to‑haves
Experience working with thermal infrared imagery, processing and analysing SAR data, or combining observations from multiple sensors. Deep experience in one modality is valuable.
A PhD in geoinformatics, remote sensing, Earth observation, or a related field.
Experience scaling geospatial analysis with Dask, Apache Spark/Sedona, or Zarr.
Experience deploying and optimising GPU inference with PyTorch/CUDA, ONNX Runtime with TensorRT, or NVIDIA Triton Inference Server, including batching image tiles and managing GPU memory and throughput.
Spatial SQL with DuckDB or PostGIS, STAC-based data discovery, and exploratory work in QGIS or geemap; producing COG and GeoParquet outputs for downstream GIS use.
Experience working with commercial imagery from providers such as Airbus, Satellogic, SatVu, or ICEYE.
Experience working in defence and intelligence environments or on related projects.
Strong interest and practical ability in agentic software development: using coding agents to plan, implement, test, and review software, and keeping up with rapidly evolving tools, techniques, and trends.

Nice-to-haves

None flagged.

Application prep

Use the role’s language and application details while tailoring your materials.

Keywords to mirror

kubernetespythonmachine learningsql

Platform: greenhouse
Effort: quick

Benefits

Benefits detected in the listing.

Vision

Professional Development