Announcing 12 AI for Science recipients: building tools that turn discovery into impact
Science moves the world forward, but discovery has been slowing. To change that, Google.org created a $20 million AI for Science fund backing academic, nonprofit, and startup teams using AI to crack hard scientific problems. The goal is simple: equip researchers with resources to do in years what used to take decades.
Today we're announcing twelve funded teams. They're not just processing data - they're building AI systems that remove bottlenecks in health, agriculture, biodiversity, and clean energy. Each team commits to open science, with data and solutions shared to push the field beyond a single lab or use case.
What this means for researchers
- Expect open datasets, models, and tools you can reuse, benchmark, and build on.
- Cross-domain transfer is a theme: protein models inform plant immunity; multiscale models link single cells to organs.
- Workflows are closing the loop between AI, experiment, and field deployment - shortening the path from hypothesis to validation.
Decoding life and health
- UW Medicine: Using Fiber-seq and long-read maps to characterize hard-to-resolve regions of the human genome, with an eye on rare disease genetics.
- Cedars-Sinai Medical Center: BAN-map, an AI-guided system that analyzes neural signals and adapts experiments in real time to study thought and memory.
- Technical University of Munich: A multiscale foundation model connecting cell-level states to organ-level outcomes, enabling in silico disease progression and treatment testing.
- Infectious Disease Institute, Makerere University: Applying the EVE framework and AlphaFold to anticipate malaria parasite evolution and spot drug resistance earlier.
- Spore.Bio (France): An AI-enabled scanner to detect life-threatening, drug-resistant bacteria in under an hour instead of days.
Strengthening global food systems
- The Sainsbury Laboratory: "Bifrost," using AlphaFold3-informed predictions to model how plant immune receptors interact with pathogen molecules and speed up breeding for disease resistance.
- Periodic Table of Food Initiative (PTFI): An AI platform mapping the "dark matter" of food - thousands of understudied molecules that drive nutrition and flavor - to inform healthier diets.
- Innovative Genomics Institute, UC Berkeley: Decoding cow microbiomes with AI to identify edit targets that can cut methane emissions from livestock.
Protecting biodiversity and planetary resilience
- The Rockefeller University: AI-accelerated genome sequencing pipelines that automate data curation and scale high-quality genomic references for ~1.8 million species.
- UNEP-WCMC: Large language models scanning millions of records to build the most complete distribution map yet for 350,000 plant species - a decision tool for conservation planners. Learn more
- Swiss Plasma Center, EPFL: A shared data standard and experimental framework so fusion research can train AI on collective results and move faster toward reliable, carbon-free energy.
- University of Liverpool: A "Hive Mind" that links autonomous lab robots, human scientists, and AI agents to discover new materials for large-scale carbon capture.
Why the open approach matters
Each project commits to sharing data and methods. That means better baselines, more reproducibility, and fewer dead ends. It also lowers the barrier for smaller labs to test ideas without starting from scratch.
For teams building on this work: document data lineage, publish preprints with clear evaluation protocols, and prioritize interoperability (formats, APIs, ontologies). That's how individual breakthroughs compound.
Looking ahead
From decoding rare disease biology to mapping plant diversity at scale, these teams show how well-aimed AI can shorten the distance between insight and real-world impact. We'll continue supporting this cohort and look for more projects with strong scientific merit, a path to deployment, and a plan to share results broadly.
If you're upskilling your lab with practical AI methods and workflows, explore curated learning paths by research role at Complete AI Training.
Your membership also unlocks:
