Offline AI Applications: Graduate from MBZUAI in UAE develops intelligent software without internet dependence
Graduate's AI Compression Project Aims to Bridge Digital Divide
Daniel Gebre, a recent graduate from Mohamed bin Zayed University of Artificial Intelligence (MBZUAI), has developed a groundbreaking project called iShrink that aims to bring offline AI tools to people without internet access.
iShrink is a framework that compresses large language models (LLMs) to run efficiently on mobile devices locally. By identifying and removing less important parts of the model and fine-tuning the remaining components, iShrink reduces the size of LLMs, making them faster, lighter, and suitable for devices with limited resources.
In practical terms, iShrink achieved approximately a 22.5% and 19.7% size reduction on models such as LLaMA 3.1-1B and Falcon 1B respectively. This means that these models, which traditionally have fewer redundancies to compress, can now run fully offline on mobile devices, allowing students and users without reliable internet to access advanced AI capabilities directly on their phones.
Gebre also developed a mobile app prototype that runs entirely on the device, demonstrating the practical offline use of iShrink-compressed models. Future plans include adding voice and multimodal features and releasing iShrink as an open-source tool to foster community use and improvements.
After graduation, Gebre joined Inception, a G42 company, as an Applied Scientist, where he focuses on domain-specific AI solutions for industry. However, his long-term mission remains expanding access to technology for underserved communities, a passion that was ignited during his research internship at MBZUAI.
Under the mentorship of Dr Moayad Aloqaily and Professor Mohsen Guizani, Gebre became immersed in the possibilities of AI and how it could address real-world problems. His master's research at MBZUAI focused on bringing AI tools to people without internet access, a project that culminated in the development of iShrink.
Gebre's advice to students from under-served communities is to master the fundamentals, no shortcuts, and to focus on strong interpersonal skills and building a solid professional network. He believes that these skills will be crucial in driving positive change in underserved communities.
Moreover, Gebre plans to return to Eritrea to work on digital infrastructure and education. He believes that the lack of digital representation of Eritrea's many ethnic groups and languages is an issue that can and should be addressed. By bringing AI tools to people without internet access, Gebre hopes to help students like him learn, explore, and grow, regardless of their location or resources.
[1] Daniel Gebre, iShrink: Compressing Large Language Models for Offline AI, Master's Thesis, Mohamed bin Zayed University of Artificial Intelligence, 2025.
[2] Daniel Gebre, iShrink: A Framework for Offline AI, Proceedings of the 2025 International Conference on AI, Dubai, 2025.
- Daniel Gebre's master's thesis at Mohamed bin Zayed University of Artificial Intelligence, titled iShrink: Compressing Large Language Models for Offline AI, presents a groundbreaking framework that aims to bridge the digital divide by allowing AI tools to run offline on mobile devices.
- In the realm of technology, Daniel Gebre's iShrink framework compresses large language models (LLMs), enabling them to run efficiently on devices with limited resources, opening up access to advanced AI capabilities to individuals without reliable internet access.
- By focusing on education and digital infrastructure, Daniel Gebre plans to return to Eritrea to help address the issue of the underrepresentation of its ethnic groups and languages, using AI tools to empower students like himself, regardless of location or resources.
- The practical application of iShrink-compressed models is demonstrated through a mobile app prototype developed by Daniel Gebre, a recent graduate from Mohamed bin Zayed University of Artificial Intelligence and now an Applied Scientist at Inception.
- The art of AI and its potential to address real-world problems was a passion that was ignited during Daniel Gebre's research internship at Mohamed bin Zayed University of Artificial Intelligence, under the mentorship of Dr Moayad Aloqaily and Professor Mohsen Guizani.
- Daniel Gebre's advice to students from underserved communities is to master the fundamentals, focus on strong interpersonal skills, and build a solid professional network, as these skills will be crucial in driving positive change in underserved communities.
