BudgetPilot

We were inspired to improve the shopping section of PayPal. Shoppers can accidentally buy beyond their budget or have difficulty finding products that match their needs, so we decided to build an AI feature that balances budgeting and buying needs. To make the shopping process easier for users, we wanted to turn their natural language into personalized, cost-focused recommendations. We developed BudgetPilot using Python, Gemini API, Streamlit, and PayPal API. Gemini processes the natural language by turning instructions into JSON, and the recommendation algorithm matches these requirements with a sample product catalog. A challenge we faced was coordinating work within our team, especially when parts depended on one another. The Gemini parser, recommendation algorithm, and PayPal integration were developed separately, so we agreed on input and output formats and dependencies. From this project, we learned that building an AI-powered tool goes beyond simply calling an LLM, as we had to combine it with budgeting algorithms and API integrations.