As artificial intelligence (AI) systems grow larger and more complex, the central challenge for researchers is no longer just to design smarter algorithms, but to also train and deploy them efficiently across vast networks of processors. For Saudi Arabia to fulfill its ambitious AI goals, it is investing heavily in AI technologies through advanced science and education programs. The Kingdom is especially dependent on King Abdullah University of Science and Technology (KAUST) continuing its growth as a leading research hub for AI innovation. 

To continue its support of AI research and development, on November 24–26, 2025, KAUST is hosting the workshop Distributed Training in the Era of Large Models. Organized by Professors Peter Richtarik and Marco Canini, the workshop will bring together international experts in computing, mathematics, and AI to discuss how to scale large models such as large language models (LLMs) and vision transformers (ViTs).  

Scaling AI as a national priority 

Saudi Arabia’s Vision 2030 aims to position the Kingdom among the top ten nations in AI research and application. Recent indicators show rapid progress, with Saudi Arabia now ranking among the top 20 countries globally for AI talent density, and AI is expected to contribute $235 billion to the national economy by 2030. 

This year, KAUST reached its highest global ranking yet and was the top Middle East institute for AI talent in the Global AI Competitiveness Index, an international benchmark that compares nations in AI competitiveness. The Center of Excellence in Generative AI and suite of AI education programs produced by KAUST Academy are just two examples of the University's growing commitment to AI. 

“KAUST offers an outstanding research environment and facilities and is full of highly talented and motivated people eager to push the horizons of knowledge, which makes the University one of the best academic institutes in the world to work on optimization, machine learning, and AI,” said Richtarik  

Solving critical bottlenecks in AI 

Richtarik’s research focuses on the mathematical foundations that make large-scale AI possible. With his team members Alexander Tyurin, Artavazd Maranjyan and others, he recently solved a 75-year-old problem in asynchronous optimization, providing time-optimal stochastic gradient descent algorithms for training a machine learning model with several parallel processors that work at different speeds. This work addresses one of the main barriers to scaling AI efficiently: the need to synchronize millions of parallel computations. Richtarik’s findings improve the speed and reliability of large model training and support applications in distributed and federated learning. Federated learning is essential for digital privacy and describes a decentralized approach to training AI models in which data remains on local devices. This design allows many systems to learn collaboratively without transferring sensitive information. Further, with his Ph.D. student Kaja Gruntkowska, who received the first Google Ph.D. Fellowship in GCC (Gulf Cooperation Council) countries, Richtarik has developed a new algorithmic framework, called the “non-Euclidean broximal point method”, which acts as a blueprint for geometry-aware design of a new generation of deep-learning optimizers. 

Canini approaches the scaling problem from the systems side. His research focuses on distributed deep learning, the process of training large models across many processors to overcome hardware limitations. This work includes balancing processing loads, reducing communication bottlenecks, and optimizing memory use, all of which are critical to ensuring that large AI models can continue to grow in capability without being limited by system design. 

The complementary work of Richtarik and Canini captures KAUST’s approach to AI: addressing the underlying problems that determine how far the field can scale. As model sizes and applications continue to expand, the University’s researchers are developing the mathematical, computational, and engineering tools needed to make large-scale AI both practical and transformative. 

By convening global experts this November, KAUST is helping define the next phase of AI and showing Saudi Arabia's global leadership in AI research and innovation.