Home Harvest is an agri-tech start-up with a clear goal: make pesticide-free, home-grown salad as easy and reliable as picking it up from the shop without the environmental cost of shipping bagged leaves across the world by lorry.
They came to us in 2025 with a vision. What they needed was a partner to help turn a promising prototype into a connected, intelligent product they could put in front of buyers, and then keep evolving it as the product matured.
Twelve weeks of intensive prototype build and six months of crop trials and continual optimisation later, we're still working together across the hardware, firmware, cloud and app and the relationship has shaped almost every part of how the product is built.
Co-developing the hardware
Working alongside the Home Harvest team, we designed a custom prototype PCB around an ESP32 microcontroller with onboard Wi-Fi. The board ties the whole smart farm together: the water pump, fan, grow light, and the temperature and humidity sensors that tell the system what the plants need. A secondary interface, buttons and indicator LEDs, gives users an at-a-glance read on what the pod is doing.
That hands-on UX work fed back into the design quickly. The indicator LEDs are genuinely useful when you're walking past the pod, but a different story when one's on a kitchen counter and you're trying to watch TV a few metres away. So together we developed a "quiet mode" that dims the LEDs and drops the fan to a quieter duty cycle on demand and worked through which signals should still be allowed to break through (a low-water warning, for instance, but not a routine watering cycle).
An intelligent, connected backend
To get the pod online, we built the cloud layer on AWS IoT Core. Each pod registers as a managed device, reports telemetry over MQTT, and pulls down new growing programmes and over-the-air firmware updates as the product evolves. The control loops themselves all run locally on the ESP32 with the fan and heater working together to hold both temperature and interior humidity within the bands each crop actually needs to thrive, and the pump running on a watering schedule synchronised with the lighting cycle. The pod keeps doing its job even when the network drops, and only pushes notifications up to the app when something genuinely needs attention, like a low water reservoir.
An app that actually knows about plants
Alongside the device, we co-developed the companion app that pairs with the pod over Wi-Fi and tracks each crop through to harvest. At the heart of it is a health-focused agritech LLM we built with Home Harvest — it helps people decide what to grow based on what they want to eat, what suits their diet, and what fits the household. It's what turns the pod from "another gadget on the counter" into a reason to keep growing.
From proof of concept to a maturing product
The original engagement was twelve weeks: a demo-able proof of concept Home Harvest could put in front of buyers at trade shows. We hit that and we've spent the six months since working with the team on the next phase. That's meant the team running crop trials to gather real growth data: temperature curves, watering volumes, time-to-harvest by crop and feeding it back into the firmware tuning and the LLM's recommendations.
Among other things, we've added heaters so the pods can keep growing in cooler-climate homes, swapped in quieter fans so they sit comfortably in a kitchen, fitted flow-rate-controlled pumps for finer watering, and tuned the lighting profiles to aid germination. The firmware, the app and the LLM have evolved alongside the hardware throughout, each release informed by what we're actually seeing in the agricultural trials.
It's the kind of engagement we like best: a clear-eyed founder team with a real product to ship, and a long-term collaboration where the hardware, the software and the AI all evolve together.
.svg)
