Lila Sciences

Lila Sciences builds AI models and automated laboratories that connect scientific reasoning to physical experiments across life science, chemistry and materials science. Its AI Science Factories combine AI, software and custom hardware in a loop: models generate hypotheses, design and run experiments, learn from results, and iterate. Scientific superintelligence is the company's stated goal.[5][6]

Lila's approach makes experimentation part of model training. Its September 2025 description combines large-scale reinforcement learning across simulations and laboratory experiments, using real-world verification to teach scientific reasoning. Hypotheses, experimental designs, results and interpretations become training tokens for successive models. The company's technical description also lists tools that agents can chain together, including molecular dynamics simulators, protein structure predictors, quantum chemistry solvers, gene editors and robotic workflows.[6][8]

The laboratory architecture supports that feedback loop. Latent Space's July 16, 2026 episode, featuring CTO Andy Beam and physical-sciences CSO Rafa Gómez-Bombarelli, describes instruments as nodes on a graph, connected by a magnetically levitating transport layer and coordinated through queue-based orchestration. Lila prioritizes flexibility and generalizability over raw throughput, retaining human work where automation does not pay.[1]

In September 2025, CEO Geoffrey von Maltzahn reported that the first AI Science Factory had completed a closed loop across hundreds of thousands of AI-driven experiments. The July 2026 Latent Space introduction reported more than 10 trillion experimentally validated scientific reasoning tokens, describing them as reasoning traces rather than sequences. These are reported quantities; the experiment count and token count measure different parts of the system.[6][1]

This cross-domain platform began in Flagship Pioneering's laboratories in 2023, before Lila's public unveiling in March 2025. Its launch research already spanned therapeutics, hydrogen-production catalysts and carbon-capture materials.[5]

Lila's founding team comprised Noubar Afeyan, Geoffrey von Maltzahn, Molly Gibson, Jake Feala, Alexandra Sneider and Ben Kompa. At the March 10, 2025 unveiling, von Maltzahn was CEO, Afeyan was chairman, and Gibson was President, Future Science. Andrew Beam served as CTO, while Rafael Gómez-Bombarelli was announced as joining as CSO of Materials.[4][5]

The launch introduced the Autonomous Science platform, designed to scale experimentation with human guidance and prospective access for life- and materials-science partners. Lila reported work on genetic-medicine constructs; discovery and validation of hundreds of antibodies, peptides and binders; non-platinum-group metal catalysts for green hydrogen; and carbon-capture materials. These were company-reported research results.[5]

The launch also announced Kenneth Stanley's arrival as Senior Vice President on Beam's AI research team. By July 2026, Latent Space identified Stanley as leading open-endedness at Lila, placing machine creativity alongside the company's work on reinforcement learning and scientific problem-solving.[5][1]

Lila's March 2025 launch reported $200 million in committed seed capital, supplied by Flagship and other investors. On September 15, it announced a $235 million Series A milestone co-led by Braidwell and Collective Global. An October extension of $115 million brought the Series A to $350 million and total capital raised to $550 million. Additional investors included NVentures, Analog Devices through Flagship's existing partnership, and IQT. Reuters reported on October 14 that the extension lifted Lila's valuation above $1.3 billion, based on company information.[5][6][7][9]ⁿ¹

The financing supported factory expansion and commercial access. In September 2025, Lila announced new hubs in Boston, San Francisco and London. Reuters reported the following month that it had signed a 235,500-square-foot lease in Cambridge, Massachusetts. That figure describes leased space rather than operational laboratory capacity.[6][9]

Lila's final Series A announcement said it was welcoming its first customer cohort. Its commercial strategy assigns later development to partners: von Maltzahn told Reuters, “We aren't going to bring molecules into clinical trials ourselves, or scale up new energy breakthroughs.” Partners and startups using Lila's platform would undertake those steps. Reuters described planned enterprise-software access to the company's models and automated laboratories.[7][9]

On June 3, 2026, Bloomberg reporters Michelle F. Davis and Rebecca Torrence reported that Lila was in talks to raise about $2 billion at an expected valuation of roughly $8.5 billion before the new money. Their unnamed sources identified CalPERS and NVentures as anchors. Lila and CalPERS declined to comment; the report described financing discussions rather than a completed round.[10]