 This is a protein. It's a tiny machine. As it spins around it captures small molecules and glues them together before releasing them and this process powers the cell. Living cells are like small factories in which many different types of proteins work together as a sort of assembly line. Converting simple inputs like sugars and oxygen into more complex outputs. Now if we can design these proteins we can have them produce virtually anything around us. This could be medicines, foods, biodegradable plastics, or the dye in your blue jeans. And these factories powered by proteins can replace conventional production processes that today are based on petrochemicals or intensive animal agriculture. And the sustainability impacts of this can be huge. The global chemicals industry for example produces over 30% of total carbon emissions. A company like Sologen is disrupting this industry by using proteins in their factories to produce fine chemicals. They've built the world's first carbon negative chemicals plant that uses no heat, produces virtually no waste. It consumes CO2 and it also has a 60% margin. We're a company called Perfect Day where I worked briefly. They produce milk protein without the cow. This process uses only 1% of blue water and emits only 3% of the carbon emissions compared to cows. Now how do we design proteins? Well proteins are a structure of chains of small building blocks called amino acids. And if we swap out one of these amino acids we can change how the protein folds, how it behaves, and what function it carries out. And protein sequences can be represented as a type of code, like character strings of amino acids that can be edited. Coding biology sounds great in theory, but there's a catch. Because even a small protein can have more possible configurations than there are atoms in the universe. And so even teams of very experienced experts have a very hard time predicting the effect of a small change, let alone how these small changes impact each other. Now a founder we work with illustrated this by saying we know that 99% of our protein designs fail. We just can't predict which 99%. You see, biology, like software, has a compile and run. But the difference is that instead of taking minutes, this process takes weeks or months and it requires years of experience. And even then the success rate is only 1%. Imagine if you could build a technology that gives biologists superpowers. Technology that could dramatically increase the number of designs that work out in the laboratory. We really believe that generative machine learning is the key to unlocking this. It's very similar to some of the machine learning models that you've seen in the media recently. For example, GTP3 is a machine learning model that can help you write better marketing copy. Or with GitHub co-pilot, you could get suggestions to improve your Python code. Or even with stable diffusion, you can have a model help you make art. Now I can paint by hand and I would need thousands and thousands of hours of training to be any good at it. But today with the help of AI, I can and indefinitely faster than Vermeer, Rembrandt, or Van Gogh. Cradle is like co-pilot for biology, where a model helps you design a better protein sequence. We are building generative machine learning models to improve protein function as well as properties. And we can use these models to check, suggest, or maybe even generate entirely new protein sequences for your biological application. For example, you might want your little protein machine to produce more of the milk that Elisa just talked about. This is something biologists call activity, like how active the protein is. Or if you're making a therapeutic protein, you might want it to be stable for longer in a human body, so it can actually reach its disease target. This is something biologists call thermostability. Cradle offers a modern web-based application and set of APIs that can be used by the R&D departments of pharma, chemicals, materials, and the food industry. Our platform provides some of the best-in-class machine learning models that were validated on industry-relevant benchmarks. Biologists can collaborate, share their designs, and publish their workflows as a template for others to use. So is it any good? I probably wouldn't be the first machine learning startup founder to claim that their models were amazing. Biologists are scientists, and so you really need to convince them with data before they are willing to try something. Therefore, we've benchmarked all of our models against industry-relevant tools and data sets, and I'd love to share some of that with you. The first task that we focused on as a team was to improve protein stability under higher temperature, making sure that the protein doesn't fall apart when it gets a little bit hotter. So we've built a benchmark of around 40,000 proteins, as well as 89 peer-reviewed empirical studies that worked on improving this property of having better protein stability, and we asked our models, could you identify what changes you would make in that protein in order to make it more stable? What we found is that, before having seen any experimental information, our models were about 20% better than the state-of-art industry-leading tool that's out there today. But as we feed the models with additional data from lab experiments, these suggestions improve even more. For most of the studies, we were seeing accuracies improved to over 90%, after only a hundred experimental data points, and for those that are not biologists, most biology teams can do 89 experiments in, let's say, a couple of weeks. Cradle is working with several design partners right now to make sure that our software is fit for purpose, and it's really incredibly cool to see companies that are building proteins that are completely designed by machine learning. Protein design used to be really hard and expensive, and we want to make it easy. Success would really mean that two kids in a garage could build a bio-based product for a couple million dollars in a few years instead of ten years and tens of millions of dollars. So basically, everybody would start pitching bugs to the VC in the audience. If you're designing medicines, materials, food, or chemicals with protein, do please hit us up, and we'd love to welcome you as a design partner. For the VCs out here, we know you don't design proteins, but you do ultimately decide what gets built. If your crypto fund is down and you're looking for a different vertical to invest in, there's a lot of really awesome synthetic biology companies out there. We'll enable them going faster, but you do have to give them some money. Thank you very much.