Building AgCane AI: An AI Assistant for Sugarcane Farmers
Sugarcane is one of Karnataka’s most important cash crops, but it’s also one of the hardest to manage well. A leaf disease can go unnoticed for days before a farmer even realizes something is wrong, and by then the damage is already spreading. Add in the guesswork most small and mid-scale farmers still rely on for irrigation timing, and you get a crop that’s far more fragile than it needs to be.
That’s the gap I built AgCane AI to close.
What AgCane AI Does
AgCane AI is an AI-powered chatbot built to give sugarcane farmers fast, accessible answers about their crop’s health. It’s designed around two core problems farmers actually face day to day:
- Leaf symptom diagnosis — a farmer can describe what they’re seeing on the plant in plain text, upload a photo of the affected leaf, or do both together for a more accurate read. The chatbot interprets the symptoms and helps identify what might be going wrong.
- Irrigation guidance — farmers can share irrigation-related data, and AgCane AI helps them think through watering decisions instead of relying purely on habit or guesswork.
The idea isn’t to replace agricultural expertise — it’s to put a first layer of that expertise in a farmer’s pocket, in a form that’s immediate and easy to use.
How I Built It
AgCane AI was built solo using Google AI Studio. AI Studio made it possible to go from an idea to a working, testable chatbot quickly — I could iterate on how the assistant interpreted symptom descriptions and photos without having to stand up a full backend from scratch.
Get Chirudevina’s stories in your inbox
Join Medium for free to get updates from this writer.
The core design choice was keeping the input flexible: some farmers will find it easier to type what they’re seeing, others will get a faster, more accurate response by uploading a photo, and combining both gives the model the most context to work with. That flexibility was a deliberate trade-off — it adds complexity to how the assistant has to reason about input, but it makes the tool usable by a much wider range of people with different comfort levels around technology.
Try It
You can try AgCane AI here: https://agcane-ai.ai.studio
What’s Next
This is a first version, and there’s a clear list of what I’d want to add with more time:
- Deeper validation of irrigation recommendations against real regional data
- Support for Kannada, so the tool is genuinely accessible to the farmers it’s meant for, not just those comfortable in English
- Expanding beyond leaf symptoms to cover more of the plant’s lifecycle
AgCane AI started as a way to explore how far a solo build in AI Studio could go toward solving a real, local agricultural problem. It’s a first step — but it’s one that’s already usable today.
