A new AI-powered platform called MOSAIC is revolutionizing drug discovery by turning vast amounts of chemical knowledge into practical, step-by-step synthesis instructions. Developed by chemists at Yale University in collaboration with researchers from Boehringer Ingelheim Pharmaceuticals in Connecticut, MOSAIC acts like a digital recipe book for chemistry, helping scientists design and build new molecules more efficiently. Chemistry research has generated millions of reaction protocols over the years, but accessing and applying this knowledge effectively has long been a challenge. MOSAIC tackles this bottleneck by using 2,498 specialized AI "experts," each trained on the knowledge of a leading chemist in a specific area—such as reaction conditions, catalyst selection, or purification techniques. This approach is similar to consulting multiple world-class chefs at once when preparing a complex dish. Victor Batista, John Gamble Kirkwood Professor of Chemistry at Yale and co-corresponding author of the study published in Nature, explained that MOSAIC transforms overwhelming data into actionable laboratory procedures. “Chemistry has accumulated millions of reaction protocols, but making practical use of that knowledge remains a bottleneck,” he said. “MOSAIC is designed to turn that information overload into something usable.” Timothy Newhouse, a Yale chemistry professor and co-corresponding author, compared the system to modern tools like ChatGPT, but for chemistry. “Just as ChatGPT helps find a new recipe, MOSAIC helps chemists find the right path to make a new molecule,” he said. The platform’s unique architecture allows it to outperform traditional AI models that rely on a single, broad neural network. Instead, MOSAIC draws from thousands of niche experts, enabling it to handle complex, diverse chemical challenges across pharmaceuticals, catalysts, materials, agrochemicals, and cosmetics. In testing, the team used MOSAIC to successfully design and synthesize more than 35 compounds that had never been made before. The system also provides uncertainty estimates—indicating how confident it is in a given recommendation—helping researchers prioritize experiments based on reliability. MOSAIC is fully open-source and built to adapt to future AI advancements. Researchers emphasize that the platform goes beyond prediction, actively guiding real-world lab work. “Chemistry has evolved from books to databases, and now to AI-guided navigation,” said Sumon Sarkar, a postdoctoral fellow and first author. “MOSAIC functions like a smart cookbook and Google Maps for chemical synthesis—turning vast knowledge into detailed, reproducible procedures with a clear sense of how likely they are to succeed.” This innovation marks a major step toward accelerating drug development and making synthetic chemistry faster, smarter, and more accessible.