This Perplexity cofounder wants to help AI breakthroughs graduate from university labs

Fast Company

23-06-2025

A team of prominent AI researchers, led by Databricks and Perplexity cofounder Andy Konwinski, has launched Laude Institute, a new nonprofit that helps university-based researchers turn their breakthroughs into open-source projects, startups, or large-scale products with real-world impact. Laude brings together top academic and industry leaders to guide promising AI research out of the lab and into the world. Its mission: help more AI ideas cross the gap from paper to product.
The effort builds on a growing belief within the AI and open-source communities that the field's biggest advances should be developed in public, not behind corporate walls. Many promising breakthroughs happen inside university labs, but often end up as research papers with no clear path to deployment. At the same time, as AI's development costs and potential rewards have skyrocketed, the need to support ambitious academic work outside of the big tech ecosystem has become more urgent.
Konwinski, who was named one of Bloomberg 's ' New Billionaires of the AI Boom, ' has assembled a high-profile board for Laude. Among its members are Google's head of AI Jeff Dean, board chairman and Turing Award winner Dave Patterson, and Joëlle Pineau, a professor at McGill University and the Quebec AI Institute (Mila), and former Global VP of AI Research at Meta (FAIR).
Laude's core goal is to replicate and enhance the university lab model used by departments like UC Berkeley's, known for foundational AI research. As a PhD student at Berkeley, Konwinski helped develop Apache Spark and later cofounded Databricks to commercialize it. That experience shaped his vision for Laude. 'I could do another company,' he says, 'but I'm honestly more interested in helping find other Databricks' and Perplexities and Linux and the internet and the personal computer.'
Laude will focus on projects in four key areas: reinventing healthcare delivery (for example, by developing an AI-powered insulin pump), accelerating scientific breakthroughs (such as visualizing black holes or discovering new materials), revitalizing civic discourse (helping voters find common ground on controversial issues), and helping workers reskill for the AI age. These are domains where AI could have significant positive impact, but where the technology's potential is still largely untapped, Konwinski explains.
Laude, a nonprofit with a public benefit corporation operating arm, will award grants to ambitious 'moonshot' projects that may take three to five years to complete. Selected projects will receive $250,000 seed grants, with the most promising progressing to multiyear research labs led by faculty affiliated with universities.
'Funding ambitious, high-impact work for long periods can give academic labs the autonomy to really identify and tackle significant societal challenges,' Dean says. 'This longer-term view can enable not just writing research papers but also creation of full-fledged working systems, open-source software to catalyze broader communities, or other forms of impact.'
In addition, Laude will support 'slingshot' projects, providing fast, low-friction grants and embedded support for individual researchers aiming to launch startups or open-source projects. This could mean tens of thousands of dollars worth of compute time, funding for PhD or Postdoc support, or embedding engineers, designers, and communicators to help bring a product to completion.
'We talk about the right resource for the right researcher at the right time in order to maximize how many more open-source breakouts and how many more companies we can build,' says Konwinski, who has pledged $100 million of his own money to fund the first round of grants.
Laude's primary value will not just be resources like talent and compute power, but guidance from people who have successfully brought technologies from lab to market.
'The academic model, when done well, can be excellent, but it doesn't necessarily have this ability to accelerate research at key points,' Pineau says. 'You need to bring in more resources, build artifacts that go beyond papers, and get them in front of users.'
A network of advisers, including top professors and industry leaders, will help shape research projects by offering insights on product launches, multidisciplinary viewpoints, and best practices for open-source distribution. Among the advisers are Databricks CEO Ali Ghodsi, Jake Abernethy of Georgia Tech and Google DeepMind, Ludwig Schmidt of Stanford and Anthropic, Kurtis Heimerl of the University of Washington, Berkeley RISElab director Ion Stoica, and researcher-professors from Caltech, University of Wisconsin, and University of Illinois Urbana.
For some researchers, Laude may provide an appealing alternative to venture capital. 'There are some projects where it's probably too risky for venture capitalists to take on,' Pineau says, while noting that not all VCs are the same. 'They tend to be a little bit shortsighted and want to see returns within a certain time frame, whereas a moonshot can tolerate higher risks.' There are also practical considerations. Some researchers prefer to keep one foot in academia, while VCs often want them to go full-time in the commercial space.
Berkeley roots
The inspiration for Laude dates back to Konwinski's days as a PhD student at Berkeley from 2007 to 2012. Patterson, then a professor in the computer science department, was instrumental in developing Berkeley's lab system. There, professors lead labs that attract PhD students and postdocs to pursue emerging fields like reinforcement learning.
'We developed this model of research labs with an opinionated style that were multidisciplinary,' Patterson says. Experts from across the university were brought in to offer fresh perspectives on the work. Labs were structured with five-year sunset clauses to encourage high-impact results.
About a year ago, after founding Databricks and Perplexity, Konwinski returned to the department with the goal of using his new wealth to give more young researchers the experience he had. At Berkeley, PhD students sometimes write 'vision papers' on controversial topics. As a student, Konwinski wrote one on the value of cloud computing for research.
Upon returning, he wanted to take on an even more ambitious subject: how to accelerate and improve the real-world impact of AI research. The result was ' Shaping AI,' a paper coauthored by Konwinski, Patterson, Pineau, and others, with input from Anthropic CEO Dario Amodei, Google DeepMind researcher and 2024 Nobel Prize winner John Jumper, Eric Schmidt, and former President Barack Obama.
The idea for Laude took shape through writing the paper. 'We recognized that a way to help shape AI's impact was to set up prizes and research labs, similar to what we did at Berkeley,' Patterson says. 'The new idea was inducement prizes like the X-Prize, and also new labs in North America to tackle big problems and improve AI's outcomes for public good.'
How Laude fits in
Laude is not exactly an incubator or an accelerator. It represents something new, with a clear 'AI for good' mission and a conscientious approach to where and how the research is done.
That starts with transparency. 'One of the requirements of this funding is to keep everything in the open,' Patterson says. 'There are not many requirements for grant recipients, but one is everything must be open source.'
Konwinski is also focused on how researchers handle both the benefits and risks of the technology they create. Returning to Berkeley, he was troubled by the polarized tone of the AI debate.
'The AI discourse has ended up a bit polarized,' he says. 'It's the accelerationists and doomers. You either pump the brakes or you're pedal-to-the-metal. That loses nuance.' Konwinski believes in a rational middle ground. 'It would be just as much of a tragedy to ignore the upsides, especially medium and near-term upsides, as it would be to ignore the catastrophic potential.'
Laude will encourage researchers to participate in public discussions about their work, partly to ensure they appreciate the weight of the decisions they are making. Too often, he says, executives like Sam Altman or Sundar Pichai lead the conversation about breakthrough technologies, not the Ilya Sutskevers and Jeff Deans who actually create them.
Getting started
On Thursday, June 19, Konwinski's voice was nearly gone after presiding over Laude's first Ship Your Research Summit the day before in San Francisco. The event brought together 70 handpicked researchers from more than two dozen universities for a day of salon-style discussions. Speakers included Jeff Dean and Dave Patterson, along with an off-record session with the Databricks founding team. Laude plans to make the summit an annual event to strengthen its community and attract new talent in computer science.
Konwinski is particularly passionate when talking about Laude's community-building role. He wants Laude to serve as an anchor for researchers with strong academic ties who believe in open source and are motivated to use AI to tackle tough problems and seize new opportunities.
'It means you put people in a room and you make them like part of something bigger than themselves,' he says. 'It's like, 'Wow, I'm with my people here who want to move humanity forward by turning research into breakthroughs.' That's special.'
Shortly after the summit, Laude announced its first major investment: $3 million a year for five years, comparable to a National Science Foundation grant, to fund a new AI-focused lab at UC Berkeley. The lab, led by a team of Berkeley's top researchers including Ion Stoica, Matei Zaharia, Joey Gonzalez, and Raluca Ada Papa, is set to open in 2027.
The final deadline for Fast Company's Next Big Things in Tech Awards is Friday, June 20, at 11:59 p.m. PT. Apply today.