Somali Smart Assistant

Inspiration

As a Somali speaker, I realized that most AI tools do not support the Somali language. Millions of Somali people lack access to smart assistants that understand their native language. This inspired me to build a solution that empowers my community through technology.

What it does

The Somali Smart Assistant is an AI-powered chatbot trained to understand and respond in Somali. It provides:

• Answers to general questions
• Somali-English translation
• Educational support
• Everyday life help via voice or text

How I built it

I used open-source LLMs from Hugging Face and fine-tuned them with Somali-language datasets. The interface was built using Gradio, making it easy to interact with the model. I also explored the use of Qloo’s API for enhanced personalization.

Challenges I ran into

• Finding clean, high-quality Somali text data
• Aligning the LLM to work well in a low-resource language
• Designing a user interface that’s simple and intuitive

Accomplishments that I'm proud of

• Successfully created a Somali-speaking assistant
• Made AI more accessible to underrepresented language speakers
• Finished the full build alone in a short time

What I learned

• How to fine-tune open-source LLMs
• Importance of language inclusion in AI
• How to build a fast prototype using tools like Gradio and Hugging Face

What's next for Somali Smart Assistant

I plan to improve the assistant's understanding and accuracy by collecting more Somali training data and refining the language model. I also hope to integrate speech-to-text and text-to-speech features to support users who prefer voice interaction. In the future, I aim to deploy this tool on mobile platforms to reach more people across Somalia and the diaspora.

Built With

• gradio
• hugging-face-transformers
• open-source-llm
• python
• qloo