Technology.
The stack behind the products: open satellite data, explainable models, grounded language models, and a pipeline that shows its state at every step.
Capability areas.
Satellite Intelligence
Process and analyse multi-spectral satellite imagery — Sentinel-2 via Microsoft Planetary Computer — into per-field crop monitoring.
Explainable AI
SHAP-based model transparency that shows exactly why predictions are what they are. Built for regulatory compliance and stakeholder trust.
Foundation Models
Geospatial foundation and boundary models — in production today, the open Fields of The World (FTW) family powers one-click field detection over Sentinel-2 imagery.
Agentic AI & LLMs
Multi-agent orchestration systems and RAG pipelines that combine large language models with domain-specific agricultural knowledge.
Cloud & MLOps
Container-based architecture — Docker images, a Celery worker pool and a gated CI lane that runs the full suite before anything merges.
Geospatial Data Engineering
Production-grade pipelines for processing, transforming, and analyzing geospatial data at scale with open-source tools.
How it works.
From satellite data to an answer that carries its own provenance.
Ingestion
Sentinel-2, SoilGrids, Open-Meteo, NASA POWER, CanDCS-U6, OMAFRA Pub 811
Processing
Vegetation indices, cloud masking, feature extraction
Models
Yield prediction, FTW boundary detection, disease classification
Explainability
SHAP attribution, honest states, abstention
Delivery
Dashboard, exports, grounded assistant
Need these capabilities?
Tell us what your data looks like and what it should decide. We reply within two business days.