Build With AI Makerere Hackathon
A weekend of building, learning, and connecting. The Build With AI Makerere Hackathon brought together developers, designers, and dreamers at the American Center Kampala. Organised by Shakiran Nannyombi and Benjamin Wakida Eldon, in partnership with Major League Hacking and GDG on Campus Makerere University.
Result
Our project MoMo Scam Checker secured 2nd place at the hackathon!
The Team: Tech Clan
- Joy Wabule
- Awany Jey Daniel
- LUBEGA Grace
- Mubiru Elton Felix
The Project: MoMo Scam Checker
What Inspired It
Mobile Money (MoMo) fraud is a pressing, real-world issue affecting millions of Ugandans every day. Social engineering scams range from distressed callers claiming a "wrongful cash transfer" to fraudsters impersonating telecom customer care agents asking for PINs. These tactics disproportionately target everyday citizens, many of whom rely on analog feature phones with nothing but USSD menus.
Smartphone users can access web resources to verify numbers, but most victims are offline. The goal was clear: build a comprehensive, community-driven registry that bridges this gap, raises awareness, and gives people a way to instantly verify suspicious numbers across all major Ugandan networks (MTN, Airtel, Lycamobile, and UTL).
How It Was Built
Frontend Architecture
Built with React and TypeScript, styled with a modern, high-contrast Tailwind CSS interface featuring a sleek, semi-transparent sticky header for smooth scannability.
Offline USSD Simulator
An interactive feature phone sandbox styled as a classic Nokia 3310. Simulates real-time USSD dial codes like *256*5# so users can practice identifying scams without internet.
MoMo Copilot AI Autofill
Gemini-powered extraction lets users paste messy scam SMS or upload screenshots. The AI instantly extracts phone numbers, risk ratings, scam categories, and telecom operators.
Community Registry
A dynamic data grid displaying active flags, specific fraud tactics, and real-time community reports so users can verify numbers before engaging.
React TypeScript Tailwind CSS Gemini API Gemini Flash USSD Simulation Google AI OCR / Vision
The AI That Powers It
The Gemini API (optimised using Gemini Flash for instant response) runs two core workflows. First, structured data extraction: when a user receives a suspicious SMS like "Y'ello! You have won 1,000,000 UGX in the promotion! Call 0772109843 to claim...", the model processes the raw text and instantly autofills the phone number, operator, and scam category into the reporting form.
Second, multimodal screenshot analysis: scammers often operate via WhatsApp chats, MoMo transaction receipts, or system alert pop-ups. Users can upload screenshots directly. Gemini's multimodal capabilities perform OCR to read text inside the image, analyse the visual layout, and parse the relevant fraud details into the form without requiring a single typed word.
The AI also acts as an initial triage layer, evaluating the semantic meaning of each incident. It maps behaviours against known Ugandan scam tactics (like the "Wrong Number Reversal" or "Wewole Loan Scam") to suggest a risk rating and generate a clean summary for the report.
What We Learned
Inclusive Design
Creating tools for a diverse demographic means designing for both high-end smartphone users and individuals using classic button phones. The simulator proved that educational tools must mimic the real user environment to be effective.
Localised Context
Fraud patterns are heavily regionalised. Tailoring the app specifically to Ugandan fraud categories (Wewole Loan scams, School Emergency scams) makes the tool instantly recognisable and trustworthy for the target audience.
Challenges We Tackled
USSD Sandbox
Mimicking state-based phone keypad inputs and custom text display updates within a standard web browser required meticulous state management to feel identical to dialing an actual shortcode.
UI Balance
Security apps can easily become cluttered with frantic text warnings. We had to cleanly place high-priority MoMo security tips alongside a QR generator and partner logos without overwhelming the visual hierarchy.
Chaotic Scam Text
Scammers intentionally misspell words or mix Luganda and English to bypass automated filters. Fine-tuning the AI autofill to reliably capture clean numbers from highly unstructured data required rigorous testing.
Certificates
Event Gallery
Beyond the Code
Beyond the podium finish, the experience was truly memorable. From the amazing venue and awesome swag to the inspiring people in the room (and yes, the snacks).
It was also great meeting in person several people I have connected with on LinkedIn but had not had the chance to meet face-to-face, including Godwin Ofwono, lailah grant, Ronnie Linslay Atuhaire, Mariah Miranda .L and Byaruhanga Daniel.
Acknowledgments
A big thank you to Major League Hacking and GDG on Campus Makerere University for organising and hosting such an incredible event. It was an amazing experience, and I am already looking forward to the next one.
And to everyone who made this weekend unforgettable: the organisers, mentors, judges, fellow participants, and everyone who stopped by our table to ask about MoMo Scam Checker. That is what the tech community in Uganda is about.