This case study is limited to information Syngenta has already made public in its own materials. No confidential, proprietary, or client-specific details are included beyond what Syngenta has published — consistent with AgCision's commitment to client confidentiality on every engagement

Every season, Syngenta plants and evaluates more than 100 million corn seeds across roughly 300 locations and over one million research plots in North America — each plot about the size of a pickup truck, each containing a unique batch of genetics.
Before this project, the industry standard for evaluating germination and early plant health — known as "stand counting" — was a walk-by visual assessment: a person walking each plot, eyeballing the stand, and recording it by hand. It worked, but it was slow, expensive (estimated at over half a million dollars a year in labor alone), and limited by what the human eye could reliably distinguish. Small but genetically meaningful differences in stand quality were easy to miss, and getting an evaluation crew to the more remote research locations could mean a five-hour round trip for a single day's data.
With standability directly tied to how a new hybrid's genetics are judged for advancement, the accuracy of this single measurement had outsized influence on years of downstream breeding decisions.
 

I led the development of what became the Plant Stand Analyzer (PSA) — a ground-based machine combining an array of laser sensors with precision RTK-GPS positioning to detect and geo-reference individual corn plants to within 10 cm of accuracy, while moving across a field at up to 10 mph. I reported into Syngenta's Automation and Engineering team under Giru Dhanasekaran, with Judd Maxwell, Syngenta's North America Market Segment Lead, as the project's key stakeholder representing the research/breeding side of the business.
This wasn't a lab prototype that got handed off — it was engineered through repeated field seasons of real-world failure and iteration: welds snapped, frames fatigued, uneven ground caused problems no bench test had revealed. Each issue meant taking the machine out of the field, diagnosing what actually broke under real conditions, and rebuilding it to hold up the next season. That loop — deploy, fail, diagnose, rebuild — is what turned an interesting sensor concept into a piece of equipment reliable enough to run unsupervised across a continent's worth of test plots, season after season.

• Scaled from roughly 9,000 plots counted in its first full season (2019) to over one million plots counted every year since
• Has counted nearly one billion individual corn plants over six years of deployment
• Multiplied a single employee's effective output by an estimated 40:1 compared to manual walk-by counting
• Proved its value in an unplanned way in spring 2020, when the PSA — still mid-development — was pressed into early service to keep evaluation crews out of shared vehicles and reduce field-team exposure during the pandemic
• Became a foundation for further innovation: current thinking within Syngenta looks at combining PSA data with drone imagery and precision-planter data to extend the information gathered on each individual seed's performance, not just whether it emerged.

This project is a good example of the kind of work AgCision was built to do: take a real, expensive, at-scale measurement problem in agricultural research, and turn it into a fielded, reliable system — not just a proof of concept.

A few things this project reflects about how I work with every client:

• Independent — brought in to solve a measurement problem, not to sell a pre-built product
• Objective — the solution that survived three seasons of field testing wasn't the first idea, it was the one the data and the failures pointed to
• Competent — from sensor selection through mechanical redesign to field deployment at scale, seeing a system through from concept to a performance validated tool running across an entire continent's worth of trials
• Confidential — this write-up says only what's already public; the details that make a client's program actually work stay with the client

If you're facing a measurement problem that's currently solved by "a person walking around writing things down" — that's exactly the kind of problem worth a conversation.