 Our paper on NIPS is about the prediction of organic chemistry reactions. Learning organic chemistry is like learning a new language and there's actually a lot of similarities. You could see an atom as a letter, a molecule as a word and also like when you put a knot somewhere in a sentence for example, this changes the meaning of the sentence completely. It's the same as you would put another functional group into a molecule. So the whole meaning of this molecule or the reaction would change. Yeah, using an AI tool we built this system that takes an architecture similar to what is used in translation from English to French or whatever to build this tool that could translate from a reacting space, so the inputs to a product space, the outputs, the most likely outcome of a reaction. And we've put together a demo where people can play around, draw molecules, put some examples together and just life predict those reactions. The cool thing about our approach is that unlike other approach we are not making a query on a database to get existing data. Behind the scene we are using a neural network so that means if we come up with molecules that the model has never seen, the model will try to make its best guess. The target audience for the app we built are organic chemists in industry as well as in research because they could use it to predict reactions that are outside the scope of their expertise. This is usually they're really trained, they have years of experience in some kind of field, but when it goes away from that field it's really, really hard to predict the reaction. And so they could be more act like as generalists and then maybe find shortcuts in the chemical synthesis route that they wouldn't have found otherwise. And finding such a shortcut somewhere could mean immense savings in time and money. There are three things that we want to improve. First, the accuracy. For now we are at 82% but it would be really nice if you reach 90%. Second, also the coverage is that for now we are limited ourselves to a reaction coming from patents and we want something broader including the work chemistry. And third, accessibility. I think it's important to make a tool that could be used by other people by chemist students so we want to make something cloud based.