It would be hard to do a weekly series on Berkeley's impact on technology without getting to @matei_zaharia. His Berkeley story starts with Apache Spark, which began as a research project on campus in 2009, when Matei was a PhD student in the lab that became AMPLab, working on ways to make large-scale data processing faster and more flexible. Spark eventually became one of the most widely used open-source systems for processing enormous datasets. Then the researchers behind it set the foundations for @databricks. Matei and a group of fellow Berkeley researchers turned the technology they had been developing on campus into a company that has since become one of the defining data and AI platforms of its generation. It's now valued at $190B! The story didn't end with Spark, either. Matei has continued working on widely used open-source data and AI projects including MLflow and Delta Lake, while Databricks has grown from a Berkeley research spinout into a company serving organizations around the world. And now there's another Berkeley chapter. Matei is still Databricks' Co-Founder/CTO, but he's also back at Berkeley as an Associate Professor in EECS, conducting research and working alongside another generation of students who may eventually spark pioneering research and/or companies of their own. Today, his research focuses on one of the biggest open questions in AI: how to build and scale reliable agents. He's a co-author on recent open-source projects including DSPy and GEPA, which automatically optimize prompts and models so agents get better at specific tasks. This April, ACM awarded him the 2025 ACM Prize in Computing for the distributed data systems that made large-scale analytics, machine learning, and AI possible. A fitting bookend: his Spark dissertation won ACM's Doctoral Dissertation Award back in 2014. I talk a lot about Berkeley's startup flywheel because Matei's career is a pretty clean example of how it works. A PhD student invents important technology on campus. The work is released openly and gets adopted around the world. A company grows out of it and becomes a category leader. The people responsible stay connected to Berkeley, and the next generation gets to learn and build alongside them. Nearly two decades after Spark got its start on campus, that cycle is still going.

.@profjsb is the kind of person this Berkeley spotlight series was made for: a professor who has published hundreds of academic papers, spent decades at the intersection of astronomy, data, and machine learning, and already built and sold an AI startup over a decade before the AI hype curve. Now he’s building @valencyio, and @thehousefund is proud to be backing him. Josh is a Professor of Astronomy at Berkeley, faculty at Berkeley AI Research Lab (aka BAIR), and is founding co-PI of the Berkeley Institute for Data Science. Earlier in his career, he co-founded Wise.io, which used machine learning to help companies make sense of large datasets and was eventually acquired by GE. Valency comes from another problem Josh has spent years around: scientific knowledge is growing much faster than any individual researcher can reasonably keep up with. Valency is building a home for people-centered, AI-accelerated research. AI can help scientists search, synthesize, and reason across enormous amounts of information, but only if those systems have reliable access to the actual scientific record and can stay current as new work is published (in a world where the rate of new papers being published is rapidly accelerating due to AI). Valency is building infrastructure to help connect AI systems directly to that body of knowledge, so researchers can spend less time digging through information and more time using it. Plenty of AI companies are being started because a new technical capability suddenly exists. Josh is coming at this from the opposite direction. He has spent his career as a top researcher and scientist, knows where researchers lose time, and understands the gap between what today’s tools can do and what scientists actually need. Berkeley has a long history of professors turning years of research and domain expertise into companies, and this is a very good example of that path. Excited to see where Josh and the Valency team take this one.