🌋 BioTales: The Arm-Native Learning Revolution


✨ Inspiration

Learning biology can be dense and overwhelming. Textbooks often fail to make complex processes engaging or memorable. BioTales was inspired by the idea that storytelling, gamification, and visual metaphors can make abstract concepts intuitive. By turning cells into characters and molecular pathways into epic adventures, we wanted to make learning biology as captivating as reading a favorite fantasy novel — all directly on an iPad, without requiring cloud servers.


🚀 What it does

BioTales transforms biology lessons into interactive stories, illustrations, and mini-games. Users select a topic, and the platform generates:

  • 📜 Narrative stories explaining scientific concepts as epic adventures
  • 🎨 Illustrations that visually bring the story to life
  • 🃏 Matching games to connect metaphors with real scientific terms
  • Adaptive quizzes to reinforce knowledge before unlocking new chapters

All of this runs fully offline on Arm-based iPads, providing real-time responses without relying on the internet.


🛠 How we built it

We combined a React Native + Swift hybrid app with on-device AI models optimized for Arm processors.

Key elements:

  • 🤖 Text generation: LLM quantized for low memory and fast inference using ONNX Runtime Mobile + Apple Neural Engine
  • 🖼 Image generation: Small diffusion model converted to Core ML, Metal-accelerated
  • 🎮 Game mechanics: Lightweight transformer for matching exercises, fully local
  • 🗂 Data structure: Structured JSON output for story, illustrations, and quizzes
  • Performance optimization: int8/int4 quantization, thread-balanced execution on Arm cores, and Metal acceleration for visuals

🧗 Challenges we ran into

  • 🐢 Model performance on iPad: Ensuring LLM and image models run smoothly without cloud support required aggressive quantization and optimization.
  • 💾 Memory constraints: Large models needed careful memory management to avoid crashes on smaller devices.
  • 🖥 UI integration: Combining React Native interactive components with Swift Core ML pipelines required careful bridging for real-time responsiveness.

🏆 Accomplishments that we're proud of

  • ⚡ BioTales runs entirely offline on iPads powered by Arm, generating stories, visuals, and games in under 1–2 seconds per chapter.
  • 🧠 Successfully implemented interactive metaphor-based learning that makes abstract biology concepts intuitive and memorable.
  • 🔧 Optimized multiple AI models for performance, power efficiency, and seamless user experience on Apple Silicon devices.

📚 What we learned

  • 💡 On-device AI is feasible for educational applications when models are carefully optimized for Arm architecture.
  • 🧩 Storytelling and gamification significantly enhance comprehension and retention of complex science topics.
  • 🔗 Bridging multiple technologies (React Native, Swift, Core ML, ONNX) can create high-performance, interactive experiences on mobile devices.

🌟 What's next for BioTales

  • 📚 Expand the topic library to cover more advanced biology, neuroscience, and chemistry concepts.
  • 🎯 Introduce adaptive difficulty and personalized learning paths based on performance.
  • 👥 Add multi-user collaboration features for classrooms and study groups.
  • ⚡ Continue optimizing AI models for even faster generation and lower power consumption on all Arm-based devices, including future iPad models.

Built With

Share this project:

Updates