Inspiration

In many parts of the world, internet access is still limited or unreliable, restricting access to education, healthcare, and agricultural knowledge. Learuma AI aims to bring the power of AI to underserved communities even completely offline.

What it does

Edumentor Helps students and individuals learn and acquire new knowledge.

MediAssist Provides evidence-based guidance, supports multiple languages, and offers clarifications when needed, while clearly stating that professional medical advice is required for diagnosis and treatment.

Agriculture Supports farmers, agronomists, and agricultural professionals in making informed, sustainable, and productive decisions.

Assistant Mode Answers questions ranging from simple facts to complex explanations across many subjects, similar to ChatGPT.

How I built it

  • Learuma AI uses Ollama with the gpt-oss:20b model by default, but it is also compatible with the gpt-oss:120b model.
  • The system is designed to run locally on modest hardware, making it accessible in low-resource environments.
  • The user interface is simple and intuitive, allowing anyone, regardless of technical background, to interact with the AI.
  • It supports user account creation and login, which is ideal for communities where a computer is shared by multiple people.

Challenges

  • Ensuring the AI provides safe and reliable guidance in sensitive areas such as health.
  • Optimizing performance on limited hardware.
  • Providing 100% offline access to the AI and its data.
  • Data persistence and RAG option with Chroma

Accomplishments that I am proud of

  • Running the model completely offline.
  • Streaming AI responses in real time to user questions.
  • Integrating memory so the AI remembers the user’s conversation history and responds appropriately.

What I learned

Creating Learuma AI demonstrated how technology can bridge gaps in education, healthcare, and agriculture when designed with inclusivity in mind. Offline AI is not only a technical challenge but also a humanitarian one.

What's next for LEARUMA AI

  • Generate Windows, Mac, and Linux executables for Learuma.
  • Partner with NGOs and organizations to distribute Learuma AI to communities in need.
  • Further optimize the roles of AI agents (Assistant ChatGPT, Edututor, MediAssist, Agriculture) to deliver even more effective results.

Built With

Share this project:

Updates