 The goal of our work is to get new battery technologies to the market faster. Knowing when a battery is going to fail has tremendous value to the manufacturer and to the consumer. For electric vehicles to become ubiquitous, they need to be able to charge quickly, they need to have low cost, they need to have long range, and they need to have a long battery lifetime. The biggest problem that exists within the battery space right now is just the time scale for development. Developing a new battery technology usually takes on the order of decades. What we've set up here in our lab at Stanford is an important step in performing high throughput optimization of battery fast charging. One of the really exciting things about this research is we can dramatically accelerate the pace of battery innovation. The funny thing with batteries is that typically when you enhance one thing, something else goes down. So if you improve the range, the longevity goes down. If you improve the longevity, the safety men go down. So this is a conflicting property that needs to be optimized at the same time. We use a machine learning model to figure out which data streams are most important for early prediction. By learning from the data, we can avoid having to perfectly understand how batteries are degrading in order to estimate their final cycle life. First, the machine learning algorithm proposes some experiments, collects that data, learns from that data, and then uses that to propose more promising protocols. I think one of the most interesting results to come out of our battery research is actually the ability to predict what's going to happen to the battery without having to go through the entire process of testing the battery to its end of life. Currently, if you're going to predict the battery lifetime for a new battery chemistry, you would then run a bunch of charge and discharge cycles. And you would charge hundreds, 500, a thousand. Basically, if you want to determine the time it takes to fail for a battery, you have to run it to failure. For example, the high throughput cycler at Stanford allowed the generation of enough data that we could, with high reliability, develop an algorithm that could predict battery lifetime. What this means to the consumer are batteries that are cheaper and longer life, whether it's for their car, whether it's for their laptop, or whether it's for their smartphone. Our goal is to make it happen.