News - How USC is training the minds behind the latest AI algorithms

Artificial intelligence may seem like magic, but behind every chatbot, self-driving car, image generator, and digital assistant is math — a lot of it. As AI transforms industries from health care to finance to entertainment, there’s growing demand for experts who understand the mathematics behind the technology. USC Dornsife College of Letters, Arts and Sciences is helping meet that need with a new Master of Science in Mathematical Data Science — a graduate program designed to train students not just to use AI, but to understand and build it. “Data science is mathematics,” said Aaron Lauda, divisional dean for the physical sciences and mathematics and professor of mathematics and physics and astronomy. “If you want to understand AI, you have to understand the math that makes it work. That’s what makes this program unique.” Program’s unique approach gets into AI’s gears Many graduate programs focus on how to apply existing algorithms: plug in data, run models, and interpret outputs. But those tools quickly become outdated, Lauda says. “By the time you finish a program that just teaches applications, the algorithms you learned might already be obsolete.” This program takes a different approach. Its goal is to give students the mathematical foundation to implement the latest algorithms, understand why they work, and design the tools of the future. Launched this fall, the on-campus program, which spans three to four semesters, emphasizes probability theory, optimization, linear algebra, and statistical modeling — all pillars of modern machine learning. Students also gain hands-on experience working with real data sets from fields such as Earth sciences, biology, chemistry and physics. “We want students to take the theory and apply it to actual problems,” says Xiaohui Chen, co-director of the program. “That experiential learning builds a portfolio they can show to future employers or PhD programs.” Chen, associate professor of mathematics, notes that the program addresses a growing need across myriad industries. “Every sector now relies on data to guide decisions,” he said. “Unlike other programs that primarily teach how to use existing code libraries, ours trains students to create new techniques and algorithms. Graduates will leave with a rare combination of advanced math and statistical skills, mixed with practical experience.” Students say that approach is distinctive. Khang Nguyen enrolled after studying data science as an undergraduate and conducting research in reinforcement learning. “I wanted a school that would let me deepen my mathematical theory background while applying it to real-world data science problems,” he said. “We solve real challenges and work on machine learning algorithms. And the … small class sizes let me work closely with professors and dive into research questions.” In addition, USC’s location in Los Angeles offers unique benefits, including access to internships and networking with local tech companies, entertainment studios, and startups. Students also benefit from the Trojan alumni network and career support, including industry speaker events and mentorship programs. Real-world experience — and career prospects Sheel Ganatra, chair of USC Dornsife’s Department of Mathematics and professor of mathematics, emphasized that the program complements broader university initiatives, citing in particular the Institute on Ethics and Trust in Computing. “There’s a university-wide effort to ensure that AI is not only powerful but ethical and trustworthy,” he said, “and this program plays a key role in that.” Beyond being prepared for careers in artificial intelligence, graduates will also be equipped for advanced research roles and doctoral study. “Right now, the AI sector and its adjacent fields are growing rapidly,” program co-director Chen said. “Market projections estimate the AI technology market could reach $1.8 trillion by 2030, and salaries for skilled data scientists average in the six-figure range. But more importantly, students gain skills that allow them to adapt as the field evolves,” helping to solidify their long-term prospects, he added. Students are already exploring ambitious research directions. Nguyen is working on reinforcement learning projects focused on bandit theory and policy optimization techniques. His classmate, Viraj Bansal, is investigating how theoretical math, including areas like representation theory, can inform machine learning models. “This program lets us experiment and explore,” Bansal said. “We’re learning to be not just implementers of AI, but creators and innovators.” That’s exactly the kind of mindset the program is designed to foster, says Stanislav Minsker, professor of mathematics and program co-director. “If you want to go beyond the traditional data science program, to understand and build AI from the ground up, this is the place to do it,” he said.