Toph is an AI-native agricultural compliance product that translates audio recordings from an iOS app into audit-ready reports across low-bandwidth farms.

Every year, farmers in California face roughly 5–10 government audits across water, pesticide, and fertilizer usage. When these audits occur, farmers are expected to have accurate and up-to-date documentation for everything that occurs on their farm, costing farmers thousands of work hours yearly just to ensure compliance. Beyond that, farm managers have to keep track of the task logs taken by each of the farm-hands daily, to ensure that no record goes missing.

To solve this problem, my team had to create a tech product that was uniquely fitting to farm conditions.

Our ICPs use gloves, making tap actions on iOS devices difficult. Most farm-hands do not speak English, meaning any voice-based action must support multiple languages, as well as heavy local accents. Farms are in large open areas, and in most cases do not have constant access to Wi-Fi.

To understand these issues more deeply, we interviewed farmers across 30,000 acres of farmland, and tested our products with almond farms in Patterson, as well as grape vineyards in Lodi to ensure that our product could support specialty crop farms all with different processes.

At its core, our product was more than just a compliance software—it was uniquely designed to also help farm managers manage their crops and employees.

By tagging voice memos with GPS, we allowed farmers to see what was happening on their farm, and where. If a PCA (Pest Control Advisor) were to record sightings of spider mites on crops in the north field, a farmer could use our software to track its spread across sighting locations.

I designed the software to support potential scaling into that problem space. Toph came equipped with two interfaces: an iOS app, and a website. The iOS app came equipped with a recorder, as well as a locally run AI model to support a confirmation step while parsing memo information. Once the recording details were confirmed, the information was sent to the farm manager on the website.

On this website, farmers had a database of recordings filterable by employee name, date, and activity type—each recording showing a GPS preview. If our managers wished to see the collective GPS recordings, they could expand the map to track their employee tasks throughout the day at scale.

In reports, farm managers would be able to customize compliant farm report templates in the app. Once a template was set, they would select recordings from their employees, and the parsed activity details would populate the sheet recording fertilizer, water, and pesticide usage levels per acre.

After testing our product across 30,000 acres of farmland, the feedback was overwhelmingly positive. In May, our product also received positive remarks from Jamie Siminoff, the founder of Ring, in USC Lavalab’s “Demo Day.”