Lila Sciences

Our very own Christopher Fussell sat down for an interview with the Accelerate Science Now podcast, State Space, to talk about the work we do at Lila Sciences and how we do it.

Iridium scarcity has long limited the production of hydrogen from renewable energy. It's one of the rarest elements on Earth. Total global production runs a few tonnes per year, so chemists have long been searching for a more abundant alternative, with little success. Our AI-directed lab designed its own experiments, ran them in a physical, automated lab, and used every result to sharpen the next round. In only three months, we screened 2,942 catalysts and landed on something researchers had long dismissed: a modified palladium composition, stable in acid for 1,000+ hours. When the results came in, our scientific leads suspected the model had gone astray. It hadn't. The system got faster as it learned. First campaign: 4 months. Second: 4 weeks. We’re finally getting a glimpse into what a closed AI loop that runs its own experiments looks like in practice. "We got an outcome we wouldn't have come up with ourselves," says Rafael Gómez Bombarelli, our chief scientific officer of physical sciences. Explore the full research in our blog https://lnkd.in/gFnd9aWY and on ArXiv https://lnkd.in/gCMsSJZC

Most machine learning does what researchers call "hill-climbing": give an algorithm a goal, and it optimizes a straight path toward it. That’s great for finding the highest peak nearby, but to notice bigger mountains on the horizon you have to look around. Science itself doesn't care about hill-climbing; it's full of serendipitous discoveries and ideas that build on diverse results. Kenneth Stanley's Open-Endedness team at Lila Sciences tackles just that.   Explore how Stanley and his team develop ways to give AI the capacity to wander, generate and judge its own ideas, and build a body of work the way a real scientific research program does. Explore here: https://lnkd.in/gcwDfcnQ

Huge congrats to Ben Kompa, co-founder and Head of AI Lab Innovation at Lila Sciences, on being named to MIT Technology Review's 2026 Innovators Under 35 list. Ben helped shape Lila's founding vision: using AI to run the full scientific method, from hypothesis to automated experimentation to the next round of discovery. Today he's building the connective tissue between our AI and the lab itself, work that's key to how we're building scientific superintelligence. Read more on Ben: https://lnkd.in/gTHPDCj8 Want to build alongside people like Ben? We're hiring across AI research, software, robotics, biology, chemistry, materials science, and operations: www.lila.ai/open-roles

Imagine a cancer therapy that could be built in months instead of years, at a fraction of the cost and works better than what's already available to patients. That's what our science team just demonstrated. Using Lila's AI-powered platform, three scientists designed a new CAR-T therapy - a treatment that reprograms a patient's own immune cells to fight cancer. In preclinical testing, it outperformed a leading therapy used in patients today, with deeper and longer-lasting results. It's part of a bigger story. Lila's AI has learned from hundreds of thousands of real lab experiments, not just scientific papers, to design medicines that potentially last longer and work better in the body. We're building the technology that lets scientists go from an idea to a life-changing medicine faster than ever before. If that excites you, we're hiring. Come build it with us: https://www.lila.ai/team Read the full story: https://lnkd.in/gPZpgiEN

CTO Andrew Beam delivering the keynote at #RaySummit2026!

"Lila is the most ambitious thing I've ever been part of." Our COO and CFO, Jawad Ahsan, sat down with MIT Sloan to talk about the path that led him here: from GE to Axon, where he helped scale the company from roughly $1B to more than $13B in market cap, to founding 10mK, and now to Lila. What drew him to Lila was the intersection of a genuinely mission-driven company and a real moment of transformation building scientific superintelligence at the earliest possible stage, where frontier AI paired with autonomous labs doesn't just analyze data that already exists, but designs and runs experiments to make new discoveries. Jawad is hiring across finance and operations, alongside open roles on teams throughout Lila. We'd love to hear from you: https://lnkd.in/gfghRg3h

We’re excited to be part of Ray Summit in San Francisco Aug 24-26. Catch our CTO Andrew Beam's keynote and stop by our booth to meet members of the LILA team. For more: https://lnkd.in/gRswH-_a

LILA has been selected to join the U.S. Department of Energy's Genesis Mission - a national effort to mobilize AI, supercomputing, and the DOE's 17 National Laboratories to double the pace of American scientific discovery. We're proud to share that three of LILA's proposals, each submitted in partnership with a national laboratory or leading research university, were selected for Phase I awards under Genesis Mission's Lighthouse Challenges: → AI-Enabled Electrochemical Refinery with Caltech and Berkeley Lab (Lawrence Berkeley National Laboratory) → Agentic AI-Driven Manufacturing of Polyethylene from Waste Carbon with Northwestern University and Argonne National Laboratory → Autonomous Identification of Semiconductor Charge Transport Physics with the National Laboratory of the Rockies. We built our AI Science Factories to run the full scientific method autonomously - increasing the speed at which ideas can be tested against reality. Genesis Mission is a bet that AI can compress the cycle at national scale, and we're grateful for the chance to pair our platform with the instruments, data, and expertise inside America's National Labs. Thank you to our partners at Caltech, Lawrence Berkeley National Laboratory, Northwestern University, Argonne National Laboratory, and the National Laboratory of the Rockies. This is just the beginning — we'll spend the next several months developing these concepts, with the opportunity to apply for larger Phase II awards to build and deploy. More to come. #GenesisMission #AI #DOE #NationalLabs #ScientificDiscovery https://lnkd.in/gY2a7Bec

The internet was AI's fossil fuel. We fracked it, and it's running out. So where does the next internet-scale dataset come from? Our answer: science itself. On the new Latent Space | AI Engineering News, Essays and Interviews podcast, our CTO Andrew Beam and Co-founder & CSO of Physical Sciences Rafael Gómez Bombarelli lay out LILA's thesis - Running the scientific method, with nature and experiments as the verifier, is the ultimate engine for training frontier reasoning models. The conversation covers our AI Science Factories - scaled verifiers that let us do post-training on reality. You know a Latent Space conversion will be good when the hosts Brandon Anderson and RJ Honicky sit down and say “Let’s go full nerd! ”👇 [episode link in comments]