This is a submission for the MLH x DEV Writing Challenge
What I Built
My teamates and I built Centsible, an AI-powered financial assistant that lives directly in iMessage and helps people make smarter spending decisions without switching between apps.
The idea came from a simple problem: it's easy to make impulsive purchases without considering upcoming bills, savings goals, or emergency funds. Centsible lets users describe their financial situation in a text message and receive personalized guidance on whether a purchase fits their budget.
I built the backend with Python and FastAPI, integrated AI-powered financial analysis, and connected the experience to iMessage using Photon/Spectrum. I also implemented privacy-focused features, including personally identifiable information (PII) scrubbing and an image-scrubbing endpoint designed to help protect sensitive information shared in messages.
One of my biggest takeaways was learning how to connect multiple technologies into a working end-to-end product. Building Centsible pushed me to think beyond the AI itself and consider privacy, real-world usability, and how to make financial guidance accessible right where people already communicate.
Demo
Partner Technologies
We used Google Gemini for AI-powered financial analysis and Photon/Spectrum to integrate Centsible directly into iMessage. I built the backend with Python and FastAPI, enabling users to get spending advice based on their available balance, upcoming bills, and savings goals. I also implemented privacy-focused features for scrubbing sensitive information. This project taught us how to connect APIs, AI, and messaging infrastructure into a functional product under hackathon time constraints.
Hackathon Experience
I attended Hacklanta, hosted by ProGSU at Georgia State University and sponsored by Cox and Adobe. I teamed up with Pamela, David, and Justin to build Centsible.We all learnt sleeping is optional when you have a hackathon deadline. The atmosphere was creative, chaotic in the best way, and full of people turning caffeine and questionable sleep schedules into actual products. I'll remember the teamwork, the debugging adventures, and watching our idea become something real. Hacklanta taught me that building software is 10% coding, 40% errors, and 50% wondering why it worked five minutes ago.
Top comments (0)