Internal project
Freshpod — Vertical Hydroponics with AI
- ~100%
- autonomous growth monitoring
- 2022
- proof of concept built
- 2023
- first LLM connected to live production data
An internal proof-of-concept (2022) to validate something genuinely hard: vertical hydroponics, with water running down vertical channels while keeping plant growth even across the whole tower — and a system close to 100% autonomous, bar loading and removing plants.
In 2023, Freshpod also became one of our first experiments connecting an LLM directly to live production data, well before that was a standard move for us.
Project stages
- 01
Challenge
Vertical hydroponics is notoriously hard to get right: water has to run down vertical channels while keeping growth even across the whole tower, with near-zero manual intervention beyond loading and removing plants.
- 02
Approach
We built a full custom stack: a FreshPi controller reading sensors and actuators (lighting, watering) plus two OpenCV cameras for growth tracking, an MQTT broker and Node-RED as the control layer, Supabase and InfluxDB for structured and time-series data, and a Grafana dashboard plus the Freshpod app as the team's day-to-day front end.
In 2023, we connected an LLM directly to that live production data — instead of digging through spreadsheets by hand, the team could ask the system plain questions about a growth experiment and get an answer, sourced from real data, in seconds.
- 03
Result
The proof of concept worked technically: even, healthy plant growth tracked automatically across towers, and a natural-language interface that out-performed manual Excel analysis. It didn't validate a standalone business case in 2022, so it's shelved for now — but it directly shaped how we build AI-connected systems for clients today.
In detail
