Two Sigma’s success depends on giving our people access to massive amounts of computing power – instantaneously, seamlessly, and securely. Our team makes it possible. We oversee three data centers supporting thousands of machines. We help people collaborate by giving them flexible access to resources across the company. And we’re growing capacity dramatically, too. Our work is a series of management challenges, filled with trade-offs and loads of interesting problems to solve.
Every day, Two Sigma pulls in thousands of data sets and generates thousands more from analysis. Most traditional data storage techniques don’t work for us because of the sheer volume, so we need to get creative. We’re building platforms that quickly collect, organize and store the data coming in, and that scale efficiently. We’re proud to make data available as soon as it’s captured rather than making people wait for batch downloads at the end of the day. Storage should never be a bottleneck.
Before Two Sigma makes a trade, our modelers may back-test the strategy behind it with a market simulator. They may evaluate how it might perform across thousands of trading days with millions of trading events – because hunches aren’t good enough. The simulator runs on our supercomputer-sized distributed computing platform. Our team’s responsible for engineering that platform – providing computing power, and tools for launching, monitoring, managing, analyzing, and visualizing simulations at scale.
Two Sigma’s modelers create forecasts that predict how markets will behave. We use those forecasts to make trades in the world’s financial markets. We also decide how to best set up our portfolio, and how long we should hold certain investments. Sometimes our models contradict each other; it’s our job to optimize everything, and decide which moves will minimize risk and costs and maximize profit for the firm as a whole. We’re hardcore problem solvers, with strong quant and programming skills.
At Two Sigma, machine learning and artificial intelligence are in everything from information clustering and extraction to market prediction. Diverse sources deliver enormous amounts of data every second and we adopt cutting-edge distributed learning algorithms to analyze and discover today to predict tomorrow. We not only focus on applied machine learning research that generalizes single-machine algorithms to distributed environments, but also on innovative ways to improve the predictive ability of existing models to foresee the world better and faster.