Ph.D. student Majid Daliri does algorithmic battle against computing bottlenecks

The National Organization for Development of Exceptional Talents (NODET, sometimes known by the Persian-derived acronym SAMPAD) is an Iranian group that oversees a network of highly selective middle and high schools, each with a curriculum aimed at cultivating the innate abilities of the country’s most intellectually gifted students.
To call Majid Daliri, now an NYU Tandon doctoral candidate, gifted, however, is an understatement; it does little to describe the sheer determination, unshakeable work ethic, and commitment to science that have defined his journey from the Iranian city of Mashhad to Brooklyn.
As a student at a NODET high school in Mashhad, Daliri participated in multiple academic Olympiads, ultimately reaching the national level, and after graduating, he entered the University of Tehran, intent upon studying computer science. (Olympiad medalists are exempt from Iran’s national university entrance exam, as they have already demonstrated exceptional capability in university-level subjects.)
Daliri was not only capable of excelling in university-level courses, but he did so while working full-time as a software engineer at Cafe Bazaar, a Farsi app store comparable to Google Play. (His undergraduate years coincided with the COVID-19 pandemic, which eliminated the need to race between school and office, since he worked and took classes remotely.)
Being a full-time software engineer and undergrad might have been challenging enough for most people, but Daliri, eager to conduct research as well, reached out to computer scientists whose work he admired to ask about the possibility of working with them. His initiative paid off: he now counts Amir Zandieh, then a postdoctoral fellow at the Max Planck Institute and currently a senior researcher at Google, and Amir Goharshady, then an assistant professor of computing and mathematics at the Hong Kong University of Science and Technology and now a faculty member at Oxford, as important collaborators and role models.
Daliri and Goharshady were among the co-authors, for example, of “Efficient approximations for cache-conscious data placement,” which tackled the issue of minimizing runtime overhead in computer programs, and he and Zandieh co-wrote “KDEformer: Accelerating Transformers via Kernel Density Estimation,” which discusses the computing bottlenecks that occur when processing data that unfolds over time.
One other researcher came to Daliri’s attention as he further immersed himself in the field: Christopher Musco, an NYU Tandon assistant professor of Computer Science and Engineering, whose work focuses on developing new algorithms capable of efficiently processing the world’s largest datasets.
When you’re considering graduate programs, it’s obviously important to work with someone whose research interests mesh with yours, but equally important to find someone who will be a caring and supportive mentor. As soon as we began talking, I knew studying with him at NYU Tandon would be a good choice for me, and while I was accepted to more than one program, it was an easy decision to come to Brooklyn after I earned my bachelor’s degree in 2022.”
— Majid Daliri
While the decision was easy, the waiting period for a student visa was not: the clearance process took a nerve-wracking three months. “I had quit my job in preparation for leaving, and there was nothing to do but wait and hope that I would find out before the deadlines for accepting another admission offer passed,” he recalls.
Now that he’s here. Daliri is working to develop algorithms with provable guarantees to enhance the efficiency, sustainability, and scalability of machine learning systems, especially in the context of deep learning and generative AI.
That work recently won him selection into the prestigious 2025 Apple Scholars in AI/ML cohort, a competitive fellowship program that supports emerging leaders in academic research with funding, mentorship from Apple researchers, and internship opportunities while pursuing their Ph.D.s.
“Majid has been an absolute pleasure to advise as a Ph.D. student, and I have already learned a ton from working with him,” Musco says. “He has great research ‘taste,’ somehow managing to repeatedly come up with algorithms that are simple and elegant, but still blazing fast. His work on vector compression and sampling stands out as an example, and I think will have a lot of long-term impact in databases, machine learning, and beyond. Majid is also really strong mathematically — I can think of multiple instances where he has truly surprised me by cracking a challenging problem with a surprisingly elegant proof. Add that talent to an incredible work ethic, and maybe most important of all, a patient, kind, and thoughtful personality, and it’s not surprising that Majid has found so many successful collaborations at NYU and beyond.”