Fourteen concrete pieces of engineering work that together build out the five capabilities. Each card sits within one capability (or shared across two), with its definition clarity, implementation difficulty, risk grade, and current maturity in the codebase.
Click any tile to focus on it. Use the filter chips to view by capability.
Interactive web app where field teams design sample plans — where to sample, in what density, across which layers — before going out. Replaces ad-hoc spreadsheets and PDFs.
The pipeline that follows a sample from the field through the lab to its result, with full provenance and batch tracking. Without it, no downstream report can credibly claim where a number came from.
The combined remote-sensing pipeline, geodata server, and spatial reasoning layer underneath the sample plan designer. Lets every product reason about location, satellite imagery, and field geometry.
An interactive, versioned scientific notebook for the R&D team to investigate data and run computation reproducibly. The single piece of work that simultaneously unlocks reproducible R&D and provides the substrate for problem-specific reasoning.
A system that collects reference corpora (e.g. soil-property distributions, crop performance baselines) on a schedule and keeps versioned comparative statistics. Lets a customer's result be expressed as "X above the relevant benchmark".
A simple, reliable way for customers to pull their processed data out of our platform in standard formats (CSV, GeoJSON, etc.). Table stakes for any technical buyer.
A thin layer over the query server so customers and partner integrations can request specific slices of their data programmatically. The integration surface for any agri-data exchange.
A blob storage for generated deliverable files plus magic-link delivery and a versioning + lifecycle store. The pipe that turns "we have a result" into "the customer has a polished report in their inbox".
A customer-facing web app where users can click around their own results — drill into a field, compare regions, swap layers. A componentised refactor of today's dashboard, tailored per product surface.
A standardisation layer that maps incoming customer data (any schema, any units) onto the MIxS biological-metadata standard. Required so every downstream product can assume a clean, consistent input shape.
An agent that ingests scientific literature, tracks claims and evidence, and keeps a categorised store of what the field knows about a given crop / soil / agronomic question. The literature scaffolding behind any companion-dx claim.
A productised capability for plugging in the bespoke scientific reasoning each customer needs (different crops, different soil questions, different intervention claims) without re-building from scratch every engagement.
An algorithm that takes soil, environmental and historical-performance data and recommends where to place a given seed variety, treatment, or intervention. Moves us from describing fields to advising on them.
An autonomous agent that walks through farm and field data the way an experienced agronomist would — flagging anomalies, hypothesising causes, drafting interventions. The most ambitious item on the board and the one with the lowest definition clarity.