 We are applying several technologies being developed inside IBM to build a platform in which allows researchers to accelerate our ability to discover. So we set out an important sustainability challenge facing the semiconductor industry. Can we develop or discover new, more sustainable photo acid generators? Photo acid generators are one of the many materials used in the manufacturing of semiconductor devices. We have tremendous numbers of talented researchers working on all kinds of technologies. The real challenge was how do we find a common language? As part of this project, what we realized is that common language is data. The nice thing about deep searches is we can use any type of NLP and image analytics that we are not really bound to a specific domain of expertise. So we have access around 25 million patterns, we then filter them out and then what we say is let's map all of these chemicals, find all of the materials and all of the packs that are mentioned and extract them as a huge list. Once I have a list of all of the materials I can say, okay, what is special? Is there maybe a certain substructure of the molecule that is special? Are there outliers? Are there things that we were not yet expecting? One of these materials was already actually made a couple of years ago with a different purpose. They were picking it basically off the shelf and doing some experiments. The properties that were coming out from the experiments were significantly better. Because of that outlier, we realized that there might be some other white space that is not yet discovered. As a result of this, what we found is a lot of very important property information is not in the publicly available literature. So we turned to Ed and the intelligence simulation team to use simulation and computation to augment our data set. We worked with the deep search team to use our modeling and simulation capabilities to fill in gaps, providing them with information where there was previously none. We used quantum mechanics to look at the optical properties of the Pag molecules. And we also ran physiochemical models to understand the toxicology and the environmental impact that these molecules may have. Each one of these simulations could take up to days on a supercomputer. With our AI-enriched simulation method, we were able to reduce the amount of simulations you needed by about 60%. The next step of the process is to use that structure property data to train a generative AI model. A generative model is a significant game changer of material design. We discussed with chemical experts of use case teams and saw that we can identify some structural rules which make Pag structure distinct from other materials. We needed to satisfy many target properties that includes energy gap, toxicity and biodegradation rate and some other properties. More than 3,000 structures were generated by running the tool for only 6 hours. We used AI-enriched simulation to understand the molecules which the generative models were producing before sending them for synthesis. With some actionable candidates as a result of the generative model, we then partnered with TAO and the IBM Xerox team who are developing autonomous labs which can automatically carry out those forward synthesis to produce the chemicals of interest. The IBM Robotics and Lab is an integration of three technologies, cloud artificial intelligence and commercial chemical automation hardware. We actually are treating chemistry exactly as we are treating a language. And the prediction of chemical reaction or the prediction of retrosynthetic pathways is similar to the translation between two different languages. We have been ingesting into the artificial intelligence model all the knowledge about chemical synthesis. It gives you flexibility in time, in space and most important a time reduction for the completion of the entire synthesis. So we've been able to go from understanding the data universe to augmenting the data set to building generative AI models to make predicted candidates and to have an autonomous robotic system which can actually create molecules of the class predicted by that model. These are tools that are meant to solve more general chemical and materials discovery challenges. We're just at the beginning. We can see what's coming and it's going to be amazing and it's really going to usher in this true kind of era of accelerated discovery.