Agricola

Inspiration

Technological innovation has often been concentrated in urban centers and among those working in STEM fields. But what about our fellow citizens in non-"techy" professions? The burdens of inefficiency and redundancy also weigh heavily on blue-collar workers—especially farmers. With the power of Agentic AI, we can begin to lift those burdens and bring intelligent tools to the hands of those who feed us.

What it does

Tech has long favored the urban and the technical—the programmer, the analyst, the engineer. But what of those whose hands till the soil, whose labor feeds the cities that tech serves? Agricola is built for them. It is an Agentic AI not for coders, but for cultivators—a tool that speaks plainly, thinks adaptively, and remembers wisely. With irrigation guidance tailored by weather and soil, weekly health reports for crops, and sustainable tips rooted in agricultural best practices, Agricola lightens the load that farmers have borne alone. It is a reminder that innovation need not stay in the tower; it can walk the fields too.

How we built it

On the frontend, we used React.js, Chart.js, and Tailwind CSS to create an intuitive and visually clear user interface. For the backend, we built a API using FastAPI, integrating the OpenWeather API for real-time weather data. The AI capabilities are powered by Intel Tiber, utilizing both Random Forest and XGBoost models for classification tasks. We also integrated Meta-LLaMA as our language model to enable natural language interaction with the agent.

Challenges we ran into

On the backend, we aimed to implement long-term memory for the chatbot by embedding user prompts and agent responses into vector representations and storing them in a vector database. However, we encountered persistent dependency issues while working with sentence-transformers, which delayed development. Given the time constraints, we made the strategic decision to deprioritize memory features and focus on delivering a stable, working MVP. We also faced integration challenges with Intel Tiber’s LLMs, and after considerable troubleshooting, opted to reserve Intel Tiber for our machine learning models while using alternative tools for the chatbot’s conversational layer.

Accomplishments that we're proud of

This was our team’s first time building an AI Agent, and we ventured into unfamiliar territory, working with tools and APIs we had never used before. Despite the steep learning curve and tight time constraints, we’re proud of how quickly we adapted, problem-solving in real-time, and turning new concepts into a functional product. Our ability to stay flexible, teach ourselves under pressure, and still deliver something meaningful is what we’re most proud of.