Inspiration
A vital component of making deltas in climate change is getting buy-in from key stakeholders, especially farmers, which involves working in their preferred modalities - while maintaining efficiencies. As such, we worked towards helping Lithos streamline sales discovery calls with farmers made over phone calls to help them get buy-in with farmers faster and more efficiently.
What it does
Our system transcribes phone call audio to text and converts that unstructured text dump into a structured format - following up with questions necessary to validate the sale as well as additional integrations to help organize logistics like weather data, all in one place.
How we built it
We first worked on creating a standardized template for the desired structured data relevant to considerations. We then wrote code to convert audio to text and a prompt that helped fill up our base template with information and generate follow-up questions from the conversation using GPT4. We then stored the structured information in Airtable and whipped up a frontend that displayed said information as well as presented information from our weather integration based on the timeline in the conversation.
Challenges we ran into
Understanding the structure of the desired output, given a small sample of conversations. Chunking and updating data table
Accomplishments that we're proud of
- Developing end2end system in such a short period of time
- Scoping structure for the problem
What we learned
What's next for AgriTalk
Automation
Log in or sign up for Devpost to join the conversation.