Agriculture & AgTech
Field trials are expensive and noisy. Spatial confounding is the number one analysis failure mode. Bayesian models fix that.
Agriculture demands models that isolate treatment effects from spatial field variability and environmental noise. Traditional A/B testing fails when every plot is different. Bayesian spatial models decompose what the treatment did from what the field was already doing.
Healthcare, pharma, and agriculture all need to explain their predictions, not just make them.
Work in This Space
Indigo Ag
Bayesian Field Trial Analysis
Spatial Gaussian Process modeling to isolate treatment effects of microbes on plant yield. Multi-year engagement from initial project to production deployment.
Syngenta
Crop Protection Dose-Response Pipeline
Bayesian code review for XC50 assay modeling, relative potency models, and hierarchical GLM binomial models. Two full SLA cycles delivered.
Trusted By
- Indigo Ag
- Syngenta
What We Solve
Field Trial Analysis
Isolating treatment effects of microbial additives, fertilizers, and seed genetics from spatial field variability using Bayesian causal models with spatial Gaussian Processes.
Crop Yield Modeling
Probabilistic yield prediction under uncertainty. Zero-inflated log-normal distributions for bad crop years. Hierarchical across farms and regions with weather covariates.
Spatial Gaussian Process Modeling
Decomposing spatial variation in yields to separate field-level signal from noise. Treatment effect plus spatial effects plus noise decomposition.
Crop Protection Research
Bayesian dose-response models for chemical efficacy and XC50 assay modeling, supporting agrochemical R&D pipelines at global scale.
Why Bayesian
Why Bayesian for Agriculture & AgTech?
- Field trials are expensive — make them count — Partial pooling across farm and region hierarchy extracts maximum signal from limited experimental data. No need for massive sample sizes.
- Spatial confounding solved — Purpose-built spatial GP models decompose field-level variation, isolating what the treatment did from what the field was already doing.
- Uncertainty is the output — Posterior distributions over treatment effects let agronomists and investors understand the range of plausible outcomes, not just a point estimate.
“The invaluable additional expertise proved instrumental in validating and improving our initial prototype. Working together for the past two years, we continued building and iterating on models for treatment effect estimation in agriculture.”