Rows of screens glow with running code, models sharpening their outputs while compilers probe new possibilities, and algorithms strain for greater speed. At the center of it all is Assistant Professor Riyadh Baghdadi, working to make artificial intelligence fast, efficient, and reliable enough for everyday use.
For Baghdadi, the challenge is clear. Today’s most powerful AI systems demand vast computational resources, often far beyond what a smartphone, a car, or even a household robot can handle. His research focuses on shrinking that gap.
“The question is how to make sophisticated AI models work on devices with very limited power,” he says. “If we can do that, we unlock entirely new possibilities for where and how AI can operate.”
Alongside his role as Assistant Professor in Computer Science at NYU Abu Dhabi, Baghdadi is a Global Network Assistant Professor of Computer Science and Engineering at Tandon School of Engineering in NYU and a Research Affiliate at MIT.
Much of his work sits at the intersection of machine learning and compilers, the tools that transform high-level code into highly efficient instructions a machine can execute. The aim is to make AI models faster to train, faster to deploy, and capable of running locally, not just on massive cloud servers. “Training these models can cost tens of millions of dollars,” he says. “If you can accelerate that process, it changes the economics of AI.”
Some of his research looks at systems people use every day. When a phone instantly sharpens a photo or applies a filter without delay, Baghdadi’s field is part of the reason. In other cases, the work underpins technologies still emerging, like driverless cars.
He is currently collaborating with Space42, a UAE-based company developing driverless vehicles, to help accelerate the AI models that allow cars to make decisions on the move. “A car cannot pause to think,” he says. “Everything has to happen in real time.”
As AI advances, the questions evolve. One of Baghdadi’s current projects explores how large language models — the type of system behind tools like ChatGPT — handle computer code. Early findings reveal where these models excel and where new approaches could unlock even greater capability. “It means there is still a lot of work to do,” he says. “There is huge potential, and addressing these gaps will open the door to the next generation of intelligent systems.”
Baghdadi’s path has taken him from Ecole Supérieure d’Informatique in Algeria to Sorbonne University in Paris, then to MIT in the US, where he completed his postdoc before joining NYUAD in 2021. The move placed him closer to home while giving him the kind of research environment he craved. “For me, NYU Abu Dhabi was the best fit,” he says. “The academic culture is familiar, the standards are high, and the community is incredibly global.”
That breadth of perspective, he adds, shapes the classroom and the research culture. Students arrive from every part of the world, often with different assumptions, expectations, and approaches to technology. “It creates an energy you do not find everywhere,” he says. “People are excited to be here. It pushes all of us.”
Looking ahead, Baghdadi sees two main priorities. One is to keep improving the speed and efficiency of AI models so they can run locally on everyday devices, from laptops to specialized hardware. The other is to study how these systems handle programming tasks and identify where new methods could help them perform even better.
“If we understand the limits, we can build better solutions,” he says. “It’s still early work, but it’s where some of the most exciting progress can happen.”
That combination of ambition and precision defines his work: practical applications rooted in fundamental questions, grounded in a global perspective shaped by years across Algeria, France, the United States, and now the UAE.
“We are entering a period of enormous change,” he says. “Our job is to build the tools that help society adapt, and make sure that the technology we develop is fast, reliable, and truly useful.”
