The machine learning team (known internally as ‘Delta Sierra’) tackles machine learning challenges at Palantir, using data science to deliver insights from customer data. We are coders, statisticians, analysts, and problem solvers. Working with organizations and industries that are entirely new to machine learning, we help customers across sectors create signal from the noise, from streamlining operations to generating new revenue streams. We embed with our deployment teams, and use our experience in the field to integrate cutting-edge machine learning techniques into Palantir’s core products.
Learn more about machine learning at Palantir
We have incredible latitude to define how we tackle machine learning projects. We aren’t asked to find Y given X — instead, we define what are Y and X. Patterns emerge in unexpected ways. Sometimes, we approach two problems in the same industry expecting them to be the same, but they end up being completely different. On the flip side, we often discover similarly structured problems between entirely different industries. Most of these problems are — at least in some way — different than anything we’ve seen before, but also have some common element that isn’t immediately obvious. I love that the problem space is constantly changing. It keeps me on my toes.
The connection between human analysis and machine learning is one of the best things about this team. Most companies are trying to use machine learning to replace analysts. We use machine learning to empower analysts. We’re working in really important areas, where decisions ultimately need to be made by humans, which means that we get to use really interesting areas of machine learning that aren’t often used elsewhere.
At Palantir, we create machine learning systems that are immediately, operationally used by people solving real, important problems. I’m currently working with an organization that has a large fraud problem, and we’re using machine learning to create a high-precision fraud-flagging system. Within weeks, we handed fraud investigators real leads based on an initial model, received positive and constructive feedback from the investigators triaging our leads, and improved the detection model based on their feedback. There’s nothing better than getting immediate feedback from users and seeing bad guys being caught using my model.