AI Chatbot for Regenerative Farming!

I am reaching out here looking for help to get started in the AI world!

**Data **

One hour a day meeting recordings in Telugu Language where farmers come together and talk about everything and anything Regenerative Farming!

How can it be trained?

It’s a raw discussion with some champion farmers as well. Many times the right answer is provided by someone in the group. Sometimes no one knows the right answers. And very rarely a wrong answers are also given.

Giving weights, Flagging and answering the un-answered questions… You might know better.

What’s the need for it?

There are many farmers who struggle to mute and unmute themselves during their first few calls, and send messages to the group’s WhatsApp. They don’t know how to operate or are illiterate. Some overcome this through voice notes.

If there is a system that’s trained with enough data. It can really help many farmers I know as of today.


P.S. I understand tech and am ready to jump and do whatever I can or find some other AI tool that you can help me familiarise to do it for me :smiley:

1 Like

hey Tejas, when you say you are a tech guy and talk about building a fundamental model in the first go - I doubt that. :slight_smile: (Don’t worry, I am pulling your legs).

Here is how you should be thinking to start.
Step-1: Identify/select one small use case that can be converted into a response-based AI chatbot.

Step-2: Interview, a minimum of 10 farmers and convert their answers into step by step process (Here you need to be careful about the response part from an actual user. Since they are farmers, keep that as shorter as you can).

Step-3: Create a sheet or Google doc with these user inputs and AI responses (that you have created after interviewing farmers).

Step-4: You can use ChatGPT or Google Bard (I like Bard because I can convert responses into individual Google docs) to actually fine-tune the steps, input, and responses.

Step-5: Once you are satisfied, use the embedded method (Context-based extraction) to feed these responses into any fundamental models - ChatGPT, LLaMa-2 (This is available and Opensourced for commercial usage) - and fine-tune your responses with existing weights.

Step-6 Test your responses with finetuned model weight. And if working apply this for the rest of the use cases that you have - the method would be the same.

Congrats! You have AI Chatbot for regenerative farming.

Here is the list of tools that you might need in this journey:

  1. True Foundry ( - the found is a really nice person always up for handholding and help.
  2. Hugging face ( - this has been helpful in our own journey.
  3. Inferless ( - New Indian startup offering serverless GPU that you might need)
  4. Llama 2 (Llama 2 - Meta AI) - Commerical applicable open-sourced fundamental LLM model with 70 billion parameters.

PS: I am yet to come across any vertical (Agriculture/farming) fundamental LLMs. But in any case - if you are serious, it will take time. :slight_smile:


Hello Tejas, your question made me remember this video I had seen some time ago-

The company behind it states that the stack can be used for various applications -

Check out their APIs page and see if it fits your use case.

I hope this helps!


Wow! I was able to upload some documents containing SOPs in Natural Farming and ask Jugalbandi API to answer my questions from these documents in audio format in Telugu, all within 15 mins of entering their website. Looks very promising, will explore this further!


@abhishektiwari response/playbooks?