Learn-LM

Learn from history's greatest minds. LearnLM lets you study with AI versions of Einstein, Curie, or Turing-voice tutors who teach in their style, review your notes, and make any subject come alive.

Comment

Inspiration

Imagine learning quantum mechanics from Richard Feynman, relativity from Einstein, or problem-solving from Batman. We set out to build the perfect tutor: endlessly patient, deeply engaging, and available 24/7. Education shouldn't be limited by time zones, affordability, or access to experts—so we created Learn-LM to democratize personalized learning.

What it does

Learn-LM pairs you with AI tutors embodying historic figures, fictional characters, or custom personalities you create. Select any SFU course or subject, and your tutor generates a personalized curriculum tailored to your learning style and pace. Through real-time voice conversations, your tutor provides specific, adaptive feedback that evolves with your progress—like having a world-class educator in your pocket.

How we built it

  • Frontend: React.js, Vite, Shadcn UI
  • Backend: Node.js with Hono framework, Cloudflare D1 (database), KV (caching), R2 (object storage), AI Gateway, deployed via Docker
  • AI Stack: ElevenLabs (voice synthesis), OpenAI (language models), Cloudflare AI (inference), Black Forest Labs (image generation

Challenges we ran into

Latency was our biggest enemy. When your entire experience depends on real-time AI generation—course outlines, personality-driven responses, adaptive feedback—every millisecond counts. We tackled this through aggressive optimization: benchmarking models, implementing smart caching strategies, and ultimately redesigning entire features to eliminate user wait times without sacrificing quality.

Accomplishments that we're proud of

  • Built an HTTP-accessible MCP server that enables anyone to integrate our personality-driven tutoring system into their own applications
  • Achieved millisecond response times between tutor and student through edge computing and optimization
  • Dynamically generate authentic tutor personalities that capture the teaching styles and perspectives of historical figures
  • Democratized access to education by providing free, personalized tutoring that adapts to each student's unique needs
  • Deployed on Cloudflare's edge network for infinite scalability and consistently low-latency experiences worldwide

What we learned

This was our team's first deep dive into Cloudflare's AI ecosystem, RAG (Retrieval-Augmented Generation), and MCP servers. We also built our first Terminal UI (TUI) tool for database seeding and system initialization—making deployment smoother and more developer-friendly.

What's next for Learn-LM

  • Dynamic video lessons: AI-generated educational videos tailored to your learning topic, created on-demand
  • Persistent user profiles: Track your learning journey across sessions and devices with comprehensive progress analytics
  • Expanded knowledge base: Integration with additional tools and data sources while maintaining our lightning-fast response times
  • Mobile apps: Native iOS and Android experiences for learning on the go
  • Collaborative learning: Study groups with multiple students and AI tutors facilitating discussions

Built With

Share this project:

Updates