Argi-mate AI

Home page UI
Chatbot
Natural Disaster Sign up

Inspiration

With global demand for food increasing, the agriculture industry needs the right tools to rise to the challenge. Lack of experience, gaps in knowledge, and inadequate disaster preparation can all be showstoppers for new farmers. Argi-mate AI takes these common barriers head on: our farming chatbot provides farmers with all the information they can ask for, and to keep them safe, we alert farmers to potential natural disaster based on machine learning analysis of weather data in their area.

What it does

Argi-mate AI pairs a new farmer with a farming tailored chatbot that will provide them with personalized, specific recommendations based on their inputted information. We also use a machine learning algorithm to analyze weather data to predict possible natural disasters in the farmer's location.

How we built it

We divided the overall platform into separate tasks of a frontend, backend, and setting up our Ollama endpoint. Each team member took on a particular portion.

Challenges we ran into

Prior to using an Ollama endpoint, our original idea was to train our own LLM based on farming research materials. However, this provided to be too ambitious a task for the 24hr timespan we had at our disposal, prompting us to switch to an Ollama endpoint. Unfortunately, this also proved to be more difficult than we initially expected, though with a thorough debug process, we were able to work through it.

Accomplishments that we're proud of

Attaching our Ollama endpoint was a great, long struggle, which made the moment we finally got it to work incredibly cathartic. Another achievement was successfully working through our version control issues, as we had a number of hiccups along the way.

What we learned

We had a reality check with our LLM training model, while we were training it, we realized just how much memory and time it would take. It gave us a practical lesson on how important a GPU is to the training process. We also got hands-on experience with various different servers to connect our backend to our frontend.