SoilFlux — The AlphaFold of Soil

SoilFlux is soil-microbiome intelligence: one global AI model of the ground beneath agriculture. Given a location and field context, it forecasts biological and chemical soil properties — and what to do about them — with calibrated confidence. (Visit SoilFlux)

As Co-Founder & Technical Lead, I own the platform's technical direction end to end — from the data and modeling stack to the serving API and the companion sensing hardware.

The Problem

60–70% of biological field trials fail to replicate, at €500k–€5M per trial. Inoculants and biocontrol agents fail to establish because of microbiome incompatibility, climate mismatch, or resistance from the native community. It is an information problem before it is a biology problem.

What the Platform Does

• Screen — identify viable candidate strains and trial sites before committing budget to a trial.
• Expand — predict persistence, registration potential, and where a product will travel.
• Explain — evidence-based answers with explicit, calibrated uncertainty rather than unqualified promises.
• Validate — iterate predictions against real trial outcomes.

Beyond Software

A companion in-field sensing unit brings the same intelligence to the ground itself — gas, environmental, and microscopy readings for real-time soil biological activity — closing the loop between prediction and measurement.

Stack

Python · FastAPI · Machine Learning · Geospatial data · Edge sensing