AI / ML/2026/Production
Araba — University AI Assistant
A WhatsApp assistant for University of Ghana students. It answers questions about fees, hostels, admissions, and the calendar in English or Twi, and every answer is grounded in the university’s own documents. The LLM and the messaging layer are both swappable, so another school could pick it up without a rewrite.
Context
Students at the University of Ghana wait on overloaded support channels for answers about fees, hostels, and admissions, and plenty of them would rather ask in Twi than in English. Araba meets them on WhatsApp, where they already are.
Approach
- 01Built a WhatsApp-first assistant on the Meta Cloud API, with an adapter layer and a local stub for offline testing.
- 02Grounded every answer in official university documents through pgvector retrieval-augmented generation.
- 03Made the LLM and embedding providers swappable (OpenAI, Anthropic, Google, Azure) so the system ports cleanly to other institutions.
- 04Answered in English or Twi, with rate limiting and session management, and designed for horizontal expansion to other Ghanaian universities.
Notable
- Bilingual (English/Twi) answers grounded in real university documents.
- Provider-swappable architecture built for reuse across campuses.
- Meets students on WhatsApp instead of another app to install.
Next project
Egg Commerce Platform