Expert Review of Bayesian Dose-Response Models for Agrochemical Research

Syngenta

Agriculture & AgTech
PyMCDataiku DSS

The Challenge

Syngenta's internal data science team had already built a modeling pipeline for analyzing potency assays — experiments that measure the concentration at which a compound achieves a target biological effect. They were doing solid work, but wanted expert eyes on their approach. Were their modeling choices sound? Could the hierarchical structure be improved? Was the code ready for production deployment? These are the kinds of questions that are hard to answer from inside a team, especially when the statistical territory is specialized.

Our Approach

We worked in a review-and-coaching format over two rounds. The team brought their models, and we brought fresh perspective. We dug into their dose-response modeling, examined how they were structuring comparisons between compounds, and reviewed the code architecture for production readiness in their deployment environment. Each round involved written feedback and live working sessions — iterative, not one-shot. The goal was always to improve their models and their modeling instincts, not to take the work over.

Results

Over the two engagements, Syngenta's team improved their models substantially. They internalized the feedback, iterated, and came back stronger each time.

By the second round, we were genuinely struggling to find things to critique — the models were well-specified, the code was clean, and the team had developed strong intuitions about where to be careful. That's the outcome you hope for in a coaching engagement: a team that no longer needs you.

PyMC Labs Team

  • Eric Ma
  • Virgile
  • Junpeng
  • Thomas Wiecki

More Work in Agriculture & AgTech

Indigo Ag

Separated real microbial treatment effects from the natural variation in agricultural fields, giving Indigo Ag a defensible measure of whether their seed treatments actually improve crop yields.

Gaussian ProcessesHierarchical Bayesian ModelsZero-Inflated ModelsSpatial Modeling

Haleon

Coached Haleon's internal data science team through the hard parts of building their own Bayesian marketing models — from missing data to GPU acceleration.

Code AuditHierarchical ModelsGPU AccelerationTeam Upskilling

Fox Broadcasting Company

Helped Fox Broadcasting's data science team build marketing models for TV advertising revenue and measure the true impact of NFL promotional campaigns.

MMMCounterfactual AnalysisBass Diffusion ModelCausal Inference

Let's Chat, We Respond Fast

Tell us about your problem. We typically respond within 24 hours.

Schedule a Consultation