Offline AI Applications: Graduating UAE's MBZUAI student develops intelligent software independent of internet connection
In the heart of the United Arab Emirates, a remarkable story unfolds. Daniel Gebre, a recent graduate from Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), has dedicated his master's research to bringing AI tools to people without internet access.
Gebre's journey began in 2019 when he moved to the UAE on a scholarship from the Ministry of Education, awarded to top engineering students in Eritrea. During his research internship at MBZUAI, he became immersed in the possibilities of AI and how it could address real-world problems.
Gebre's groundbreaking work, the iShrink framework, enables artificial intelligence tools to work efficiently without internet access. By making large language models significantly smaller and more efficient without much loss in performance, iShrink bridges the gap for people in areas with poor or no internet connectivity, allowing them to access advanced AI applications without relying on cloud servers or continuous network access.
The iShrink framework achieves this by identifying and removing less important parts of a large AI model and fine-tuning the remaining components to maintain strong functionality. This process results in models that are faster, lighter, and easier to run on devices with limited computing power, such as mobile phones in low-connectivity or offline environments.
Specifically, iShrink has demonstrated size reductions around 20-22% on popular AI models like LLaMA 3.1-1B and Falcon 1B, enabling these models to operate locally without needing constant internet.
Gebre has also developed a mobile app running entirely on local devices, showing promising results for offline AI interactions. The framework currently supports several open-source architectures and aims to expand, potentially including voice and multimodal features, enhancing usability in real-world low-connectivity scenarios.
Before embarking on this AI journey, Gebre completed his undergraduate degree in information technology at Zayed University with a focus on cybersecurity. Mentored by Dr Moayad Aloqaily and Professor Mohsen Guizani, he decided to pursue his master's at MBZUAI.
Looking ahead, Gebre plans to return to Eritrea to work on digital infrastructure and education. His long-term mission remains expanding access to technology for underserved communities. He also believes the lack of digital representation of Eritrea's many ethnic groups and languages is an issue that can and should be addressed. In response, he plans to launch an initiative to bring together individuals with backgrounds in AI, machine learning, and natural language processing to contribute to open-source projects that develop models aligned with Eritrea's diverse languages, cultures, and values.
Gebre is grateful to the UAE for giving him the opportunity to study and change the course of his life. He advises students from backgrounds similar to his to master the fundamentals, no shortcuts, and emphasizes the importance of a right mindset, consistent effort, strong interpersonal skills, and building a solid professional network. With his iShrink framework, Daniel Gebre is not just bridging the digital divide; he's shaping the future of AI for everyone.
[1] Daniel Gebre, iShrink: An Efficient Offline AI Framework for Low-Connectivity Scenarios, Master's Thesis, MBZUAI, 2025.
- Daniel Gebre's groundbreaking research at MBZUAI focuses on bringing AI tools to people without internet access, a concept developed in his master's thesis titled "iShrink: An Efficient Offline AI Framework for Low-Connectivity Scenarios".
- By making large AI models smaller and more efficient, the iShrink framework allows people in areas with poor or no internet connectivity to access advanced AI applications without relying on cloud servers or continuous network access.
- iShrink achieves this efficiency by identifying and removing less important parts of a large AI model and fine-tuning the remaining components to maintain strong functionality.
- iShrink has demonstrated size reductions of around 20-22% on popular AI models like LLaMA 3.1-1B and Falcon 1B, enabling these models to operate locally without needing constant internet.
- Gebre has developed a mobile app running entirely on local devices, demonstrating the potential for offline AI interactions.
- In the future, Gebre aims to return to Eritrea to work on digital infrastructure and education, and to launch an initiative to develop open-source projects that address Eritrea's language and cultural diversity.
- Gebre encourages students from backgrounds similar to his to master the fundamentals, emphasizes the importance of a right mindset, consistent effort, strong interpersonal skills, and building a solid professional network.
