DEVLOG #5

Jun 6, 2025

ai farming devlog

This devlog is focused on KrishiAI and it’s pivot.

I had worked on KrishiAI as my undergrad major project. I realised that I was building something wrong after building a set of features for the app which include

This made me think what problem does the app really solve?

My initial hypothesis

My initial hypothesis was that with better accuracy of vision models and accessibility provided by voice AI interface (with multi-lingual support), the farmers would find it useful.

I had finalised that the following were the most pressing issues for farmers in India

  1. Lack of access to information (especially about government subsidies and schemes)
  2. Unable to identify plant disease at the earliest
  3. Low literacy causing issues with understanding the dosage of pesticides

I had talked to around 4-5 farmers withing my network to understand their problems, but there is something that I missed.

What went wrong

The core issue was that I did not ask the right questions to the farmers. I was imposing a solution that I cherished and wanted to construct the problem around it — it should have been the other way around.

Another core issue was that I did not approach the problem with a first principles thinking mindset and forgot about the second and third order effects of my product decisions.

Apart from this I was trying to learn app dev on the go (with the help of AI) and vibe coded the whole app. This is not a bad thing but it made be more app focused rather that solving the core problem.

Pivot

When we think about Indian farmers and Agriculture in India, the problems are systemic and need to be solved by state level actors. Not every problem can be solve by an app but there are somethings that should make the average farmers life easier.

How can we reimagine an agriculture app that is enganging and solves a problem so well that people spread it with word of mouth.

I have a few ideas that need to be tested with real usage by farmers. These ideas have retention built-in and rely heavily on notifications and alerts.

Raw data until you process and present it in a meaningful way for the user to consume. So here are some ideas that I planning to pursue and check the waters -

  1. Crop Schedules
  2. Pest and Disease Alerts
  3. Post-Harvest Pricing decisions

Now the only work left to do is to figure out a way to present this to the farmer in a useful way. To understand the problem even deeply, I will be taking to FPO (Farmer Producer Organisation) managers to understand core problems and see if the usecases I have identified fit their scenario.