Manually tracking personal expenses is tedious, causing many people to neglect their finances. We created Budget Bot to solve this: an innovative mobile app that makes managing spending effortless. Users can simply talk to the bot or scan a receipt, and the app uses a combination of AI models to automatically extract and log the expense data. Budget Bot then intelligently assigns each purchase to the correct category, providing clear, visual summaries that empower users to understand and control their finances without the manual work.

Rapid Prototyping and Development:
The primary goal was to move from a simple concept to a fully functional, AI-powered mobile application in an aggressive 4-week timeline. This required a highly focused strategy and a streamlined, agile development process from start to finish.

Ensuring Accurate, Flexible Expense Categorization:
The AI needed to be smart enough to understand that "latte," "flat white," and "coffee run" all belong in the "Coffee Shops" category. The core challenge was to engineer prompts that could interpret natural, varied user language and consistently assign expenses correctly.

Creating an Effortless Input Experience:
To truly eliminate manual entry, the app had to seamlessly handle two very different inputs: unstructured voice commands and complex data from receipt photos. This required integrating separate AI models for speech and vision into a single, intuitive user interface.

A Focused MVP and Agile Workflow:
We strictly defined the Minimum Viable Product around the core features of voice/receipt capture and AI categorization. By working in an agile, iterative manner, our team was able to design, build, test, and deliver the working mobile application within the ambitious 4-week deadline.

Advanced, Context-Aware Prompting:
We developed a sophisticated prompt engineering framework that provides the language model with user context and a flexible understanding of spending categories. This allows the AI to accurately classify a wide range of naturally phrased inputs and ensure reliable financial summaries.

Integrated AI for Voice and Vision:
We used native iOS speech recognition for fast voice transcription and Amazon Rekognition to reliably extract text from receipt images. These specialized AIs feed structured data to the core language model, creating a powerful system that allows users to log any expense in seconds.