Landing page where you can run instant predictions with sign-in button above to access sign in and sign up pages
Input features required from farmers for making crop yield predictions
Crop selection dropdown menu for choosing different crop types
Attempted instant prediction from the landing page without signing in, and this is the response
AI-generated recommendation result based on the prediction output
Downloadable prediction receipt generated after a successful prediction
Sign in page for existing users
Sign up page for new user registration
User successfully signed in, displaying loggedIn account (email shown at the top). Navigation includes AgriBot, History and Sign Out options
AgriBot interface for AI-powered farming assistance
Example of a question asked to AgriBot and its response
Prediction history page showing all previous user predictions
Additional page displaying more prediction history records
Rainfall Impact of Yield
Average Yield by Crop and Prediction Distribution
Yield Trend over time

AgriYield

Inspiration

In many rural farming communities, decisions about planting, irrigation, and harvest timing are still made based on tradition, intuition, and past experience rather than data. While this knowledge is valuable, it often struggles to keep up with changing climate conditions, soil variability, and unpredictable weather patterns.

We wanted to build something that helps bridge this gap between traditional farming knowledge and modern data-driven agriculture. The goal was to give farmers a simple way to understand what their land could produce before harvest, using machine learning to turn environmental data into meaningful predictions and guidance.

We also wanted the system to go beyond just showing numbers, and instead provide clear, AI-generated recommendations that can help improve decision-making in real farming scenarios.

What it does

AgriYield is a machine learning–powered crop yield prediction system that estimates expected agricultural output based on environmental and farming inputs.

It allows users to:

Input farming parameters such as rainfall, temperature, pesticide usage, and crop type
Receive instant crop yield predictions
View AI-generated farming recommendations based on results
Access prediction history tied to their account
Download prediction receipts for record keeping
Explore a visual Dashboard with charts showing yield trends, crop comparisons, and rainfall impact analytics

The system converts raw model output into practical insights that can support better agricultural planning and decision-making.

How we built it

We built AgriYield as a full-stack machine learning system using a frontend web application, a FastAPI backend, and a deployed ML model hosted on Render.

The system works by:

Collecting user input through a web interface
Sending requests from the frontend to a FastAPI backend
Processing the data using a trained machine learning model
Returning predicted crop yield results to the frontend
Generating AI-based recommendations from the prediction output
Storing user predictions in a database for later access
Rendering interactive charts on a dedicated Dashboard page for trend analysis and comparison
Allowing users to download structured prediction receipts

We implemented authentication so each user has a personal workspace where their prediction history is stored securely and can be accessed anytime.

Challenges we ran into

One of the biggest challenges was connecting all parts of the system into a smooth and reliable pipeline. Making sure the frontend, backend, machine learning model, and database all communicated correctly required careful debugging and system design.

We also faced challenges with:

Handling API communication between frontend and FastAPI backend
Managing asynchronous requests and preventing delays or failures
Deploying and maintaining the ML model on Render in a stable environment
Ensuring consistent prediction behavior in production
Designing an AI recommendation layer that produces useful, grounded insights
Building interactive data visualizations for the Dashboard that are both informative and performant

Accomplishments that we are proud of

Built a fully functional end-to-end machine learning application
Integrated AI-generated recommendations with prediction results
Implemented secure user authentication and personalized history tracking
Successfully deployed a FastAPI backend for real-time inference
Created a smooth workflow from data input to actionable output
Enabled downloadable prediction receipts for transparency and record keeping
Added a Dashboard with yield trend charts, crop comparisons, and rainfall impact visualizations using Recharts

What we learned

Through this project, we learned more about:

Deploying machine learning models using FastAPI
Building and hosting backend systems on Render
Full-stack integration of ML applications
Designing reliable APIs for real-time systems
Implementing authentication and user data storage
Translating machine learning outputs into real-world usable insights
Creating data visualizations with Recharts for agricultural analytics

We also gained a deeper understanding of how important system integration, scalability, and user experience are when moving from a model to a real-world application.