Inspiration
Farms run on timers and calendars instead of data. Greenhouses run lights and air conditioning on fixed schedules completely blind to the fact that electricity costs three times more during certain hours of the day, and outdoor farms irrigate on a calendar instead of actual soil conditions and weather forecasts. We wanted to build the tool that makes both of those systems price-aware and weather-aware in real time, because the data to do this already exists publicly, nobody had stitched it together yet.
What it does
GreenWatt is a two module agricultural intelligence platform applying to indoor and outdoor farming systems. The indoor module watches live electricity pricing and shifts the greenhouse's lighting and HVAC schedule to the cheapest hours without affecting plant growth. The outdoor module divides a farm into zones, pulls real soil moisture and weather data for each one, and generates a specific irrigation and crop-health action plan. An AI Copilot sits on top of both, answering direct questions like "why is this zone stressed?" and generating daily prioritized recommendations, yield forecasts, and pest risk assessments in plain language.
How we built it
It's a single Python Streamlit application pulling three real data feeds: the National Weather Service for live temperature and solar irradiance, Open-Meteo for real soil moisture and evapotranspiration data, and a public utility tariff schedule for electricity pricing tiers and demand charges. The optimization engine calculates, hour by hour, the cheapest valid energy schedule that still meets a crop's light requirements, and a parallel calculation determines irrigation needs per zone based on real soil and rainfall data. It takes those live computed numbers and turns them into natural language recommendations and conversational answers, so the AI reasons over real data instead of generating generic advice.
Challenges we ran into
Getting the savings numbers right was harder than building the dashboard — our first water-savings calculation was off by 5x because we multiplied a seasonal irrigation need by 365 days instead of the actual growing season length, and used a water cost that was 8 times too high. We caught it by manually auditing every formula against published agricultural and utility data sources before trusting any number.
Accomplishments that we're proud of
Every number on the dashboard is calculated live from real formulas, not hardcoded, change the farm size and every figure recalculates instantly and consistently across the whole app. We're proud that the headline savings and water numbers trace back to citable public sources rather than estimates we made up to sound impressive, and that the AI Copilot is a genuinely working conversational interface grounded in the platform's live data.
What we learned
The gap between "looks impressive" and "is actually correct" is where most of the real work lives; it's easy to build a dashboard with numbers, but harder to make sure they are accurate.
What's next for GreenWatt
We can curate GreenWatt by fine tuning the AI for better recommendations. We can also expand and test GreenWatt at actual farms and expand it to further develop the app. We would also like to actually be able to implement agentic mode on the greenhouse which uses weather and electricity pricing to make decisions on its own on the lighting and HVAC in the greenhouse.
Log in or sign up for Devpost to join the conversation.