Team project · Dec 2025 – Jan 2026
CogniAdapt: an adaptive learning platform on FastAPI
AI lessons that adapt to how each student learns
- Status
- Team project
- When
- Dec 2025 – Jan 2026
- Role
- Backend developer, with Brendon on the backend and Vicky on the front end
- Stack
- Python · FastAPI · MongoDB · Groq · Azure OpenAI · React
What it is
CogniAdapt is an AI learning platform for students with learning differences, such as dyslexia or ADHD, or a language barrier. It breaks a topic into a graph of concepts, each building on the ones before it, and adapts the lessons to the learner. When a student struggles, it steps back to repair the concept they are missing instead of repeating the same lesson. It tracks progress through the graph and can read lessons aloud. We built it for a Microsoft competition.
My part
I worked on the FastAPI backend. I added a second stack alongside the original Azure services: Groq for generating the concept graphs, and MongoDB for storage. Each falls back in turn, Groq to Azure OpenAI and MongoDB to Azure Cosmos DB, then to a mock mode, so the app still starts when a service or key is missing. I also did a security pass, moving secrets out of the code and writing security and deployment guides for the team.
Why FastAPI
The backend is mostly AI calls and database reads, so FastAPI’s async support and automatic API docs fitted well. The front end is React with Tailwind, built by Vicky.