MoldLearn

Inspiration

The inspiration for MoldLearn comes from an unexpected place: the "Mold" (Megamycete) in the Resident Evil universe. In the game, the mold is a biological network that connects consciousness, absorbs information, and spreads it rapidly across a hive mind. We thought: What if we could harness this concept of a "living, spreading network" for education? In traditional classrooms, knowledge is often siloed. A student might be brilliant at Math but struggle with History, while another excels at Literature but fears Code. Just because a student is weak in one subject doesn't mean they lack the capacity to learn—they often just need the right perspective. MoldLearn was built on the belief that knowledge should spread like spores—transferring the strengths of one student to cover the weaknesses of another, creating a collaborative, self-sustaining ecosystem of learning.

What it does

MoldLearn is a collaborative, AI-powered learning platform where knowledge is packaged into micro-lessons called "Spores". Create & Spread: When a student masters a concept, they create a "Spore" (a bite-sized explanation). This Spore is then delivered to peers who are currently struggling with that exact topic. AI Personalization: Not everyone learns the same way. When a Spore is delivered, our AI engine automatically rewrites and adapts the content to match the receiver's specific learning style (Visual, Audio, or Example-based). Exploration Paths & Matching: Students can generate AI-guided learning paths. Our recommendation engine calculates the overlap between students' strengths, weaknesses, and interests to suggest peer-to-peer "Exploration Matches," encouraging collaborative learning. Knowledge Graph: Every successful Spore delivery and peer match forms a node in a visual Knowledge Graph, showing exactly how information is spreading across the classroom space.

How we built it

We engineered MoldLearn with a focus on clean architecture, scalability, and seamless AI integration: Backend: Built with .NET 10 (ASP.NET Core) using a strict Clean Architecture pattern. We utilized CQRS with MediatR to keep business logic decoupled and maintainable. Database: PostgreSQL handled via Entity Framework Core, storing complex relationships between Spaces, Members, Spores, and Knowledge Graph edges. AI Integration: We integrated OpenAI's GPT models to power our AiService. The AI handles content rewriting (adapting Spores to learning styles) and generates dynamic exploration suggestions. To optimize API costs and latency, we implemented a robust Redis distributed caching layer for AI responses. Infrastructure: The application is fully containerized using Docker with multi-stage builds, and we set up GitHub Actions for CI/CD to automatically build and push images to Docker Hub. We also integrated Cloudinary via webhooks to handle media uploads (like audio Spores and avatars).

Challenges we ran into

AI Response Consistency: Forcing LLMs to return strict JSON for our recommendation engine was tricky. We had to implement custom markdown-stripping logic and robust error handling to ensure the app wouldn't crash if the AI hallucinated formatting. Performance Bottlenecks: Initially, adapting Spores for a large classroom required sequential AI API calls, which caused severe latency. We solved this by refactoring our SporeDeliveryEngine to execute parallel asynchronous tasks (Task.WhenAll), drastically reducing delivery time. Complex Data Relationships: Designing the schema to track how knowledge transfers (Spore Deliveries) while simultaneously tracking peer collaborations (Exploration Matches) required careful planning of our Knowledge Graph entities to avoid circular dependencies.