BeSMART Recap: Teaching students how AI works
As artificial intelligence becomes capable of increasingly complex analytical and decision-making tasks, UC Berkeley is preparing young learners to do more than simply use the technology. This summer, the Department of Industrial Engineering and Operations Research (Berkeley IEOR) welcomed high school students to campus for the inaugural Berkeley Engineering Summer Machine-learning & AI Research Training, or BeSMART, program. Developed with the Dado and Maria Banatao Center for Global Learning and Outreach from Berkeley Engineering (GLOBE), the two-week residential program invited participants to look beyond the interface of emerging tools and build the programming, data analysis and mathematical foundations needed to understand how they work.
That foundation shaped the program’s progression from introductory coding to more advanced applications. Phillip Kerger, a professor in UC Berkeley IEOR, led the course with support from IEOR undergraduate Kenny Wongchamcharoen. Students progressed from Python fundamentals to working with datasets in Pandas and applying machine-learning methods in Python before exploring large language models, agentic coding and more advanced topics in industrial engineering and operations research.
During the second week, Berkeley IEOR faculty members brought their own areas of expertise into the classroom: Anil Aswani led a session on clustering, Ying Cui on optimization, Huiwen Jia on supply chains and Thibaut Mastrolia on financial markets. Together, the sessions broadened participants’ understanding of how machine learning, optimization and data-driven methods can be applied to complex real-world problems. Rather than treating AI as a black box, the course encouraged participants to examine data, evaluate outputs and connect emerging capabilities with the analytical methods behind them.
The second week of BeSMART also brought industry perspectives into the classroom through guest lectures from Sanchit Ram Arvind, a software engineer at Arize AI, and Monika Voutov, founder of Rhea TSS and Tutorkit.ai. Both are graduates of UC Berkeley’s Master of Analytics program, and their sessions gave participants a closer look at how skills in analytics, machine learning and AI can translate from the classroom into emerging technologies, startups and industry practice.
Students then put those ideas into practice through projects spanning wildfire prediction, machine learning-based agriculture management, algae bloom prediction, crime-aware map routing and the development of an efficient large language model from scratch. “I was genuinely so impressed by their projects,” Kerger said.
The experience extended beyond project work as well: Participants lived on campus, interacted with Berkeley undergraduate and graduate students and visited Google, where they gained a closer view of how AI, machine learning and computing are developed and applied in industry. By connecting foundational skills with independent inquiry, mentorship and real-world exposure, BeSMART gave students the opportunity to move from simply using new technologies to beginning to imagine how they might shape what comes next.