The knowledge engine
beneath every soil.
A global agricultural intelligence platform that unifies every analysis, research paper, extension note and field trial into a single, explainable knowledge graph — continuously learning from every sample processed by Lambda.
Search knowledge, not PDFs.
Every question below is a live capability — semantic search across millions of soils, laboratory methods and agronomic decisions.
“”
“Across Australia, what recommendations do agronomists make on sandy soils with Olsen P of 18 mg/kg?”
“Which nutrient deficiencies are consistently reported in basalt-derived soils in tropical climates?”
Eight stages, from raw document
to explainable intelligence.
- STAGE 0
Data Collection
Government archives, universities, laboratories, consultancies, farmer uploads, satellite products — every soil document, everywhere.
- STAGE 0
AI Document Understanding
OCR and LLM extraction of samples, methods, values, recommendations, reasoning and limitations.
- STAGE 0
Soil Knowledge Graph
Location → Soil Type → Crop → Method → Nutrients → Recommendation → Observed Response.
- STAGE 0
Geospatial Engine
Rainfall, temperature, elevation, geology, remote sensing and digital soil maps overlaid on every sample.
- STAGE 0
Recommendation Intelligence
Explainable AI — not what to do, but why recommendations differ, and under which conditions each holds.
- STAGE 0
Pattern Discovery
Where is sulphur deficient? Which laboratories use different critical values? Which recommendations produce measurable response?
- STAGE 0
Predictive Models
Machine learning that forecasts deficiencies, toxicities, fertiliser response and expected yield — with confidence.
- STAGE 0
Lambda Integration
Every ICP, FTIR and spectral analysis strengthens the graph — a compounding, proprietary data advantage.
Every report becomes a relationship.
Reports are not stored as files — they are decomposed into entities, methods, measurements and outcomes. The result is a living Neo4j knowledge graph in which every new sample reinforces, contradicts or enriches what came before.
- PostgreSQL — reports & metadata
- PostGIS — spatial layers
- Neo4j — recommendation graph
- OpenSearch — full-text
- Qdrant — semantic embeddings
- Object store — source documents
One graph, drawn from everywhere soil is measured.
Every document carries country, region, GPS, climate, crop, sampling depth, laboratory, methodology, author, year and confidence — metadata that makes the corpus trustworthy.
Government & Institutional
- Australian Soil Resource Information System
- CSIRO & State agriculture departments
- USDA · FAO · ISRIC SoilGrids
- European Soil Data Centre
Research & Field
- University theses & journals
- Extension notes & consultancy reports
- Mining rehabilitation & carbon projects
- Historical soil surveys
Lambda & Client Ecosystem
- ICP-OES / ICP-MS analyses
- FTIR & field spectroscopy
- Agronomist & farmer uploads
- Sensor and satellite streams
The laboratory is the flywheel.
Upload an ICP result. The platform identifies the soil, matches it against millions of comparable samples, and returns similar reports, likely deficiencies, carbon potential, nutrient density, regional trends, expected yield and benchmark farms — each backed by primary evidence.
- 1Analysis ingested from Lambda IQ
- 2Soil identified & spatially resolved
- 3Similar soils retrieved across the graph
- 4Explainable recommendations returned to client dashboard
Every discipline that measures soil.
One knowledge graph, many vantage points — each user sees the corpus through the lens of their practice.
Laboratories
Turn every ICP, FTIR and wet-chemistry run into a compounding data asset. Deliver richer client reports with contextual benchmarks.
Agronomists & Consultants
Compare paddocks against similar soils regionally and globally. Justify every recommendation with primary evidence.
Research & Government
Query decades of historical surveys, trials and extension notes through a single semantic interface.
Carbon & Rehabilitation
Track soil carbon trajectories, benchmark rehabilitation outcomes and evidence project claims with laboratory-grade provenance.
A deliberate path to a global intelligence platform.
| Phase | Focus | Duration | Outcome |
|---|---|---|---|
| 1 | Foundation | 2–3 mo | PostgreSQL / PostGIS, accounts, document storage, base map. |
| 2 | Document Ingestion | 2 mo | OCR, PDF parsing, metadata extraction, upload workflows. |
| 3 | AI Extraction | 3 mo | Structured extraction of data, recommendations, methods & reasoning. |
| 4 | Knowledge Graph | 2 mo | Neo4j linking soils, crops, methods, recommendations & outcomes. |
| 5 | Search & Analytics | 2 mo | Semantic search, geospatial queries, dashboards, filtering. |
| 6 | Recommendation Intelligence | 3 mo | Explainable AI, similarity, evidence ranking, confidence. |
| 7 | Predictive Models | 4–6 mo | ML for nutrient recommendations, yield response, deficiency prediction. |
| 8 | Commercial Platform | Ongoing | APIs, enterprise integrations, subscriptions, benchmarking. |
Straight answers about the platform.
- How is this different from a soil database?
- Databases store rows. AGI stores relationships — every measurement is linked to its method, laboratory, soil type, climate context and observed agronomic outcome, so questions about causation become answerable.
- Is my client data used to train shared models?
- No. Client data remains isolated by tenant. Only explicitly contributed datasets, or de-identified statistical aggregates, ever inform cross-tenant intelligence.
- How are recommendations made explainable?
- Every recommendation carries an evidence chain: the samples, methods, critical values and prior reports that informed it, along with a confidence score and the option to drill back to primary sources.
- Can we integrate our existing LIMS or dashboards?
- Yes. The platform exposes REST and GraphQL endpoints and can ingest results directly from Lambda IQ or third-party laboratories via signed webhooks.
Build the definitive record of the world's soils — with Lambda.
Early partnerships are opening for laboratories, agronomy practices, research institutions and carbon project developers. Contribute data, gain intelligence.