As we enter a “post Moore’s law” era,  the ability of systems and infrastructures to produce data far outpaces the ability of our systems, algorithms, and people working with the data to process it meaningfully to extract actionable insights for decision making in complex environments. While many advances in “Big Data” and new hardware accelerators are promising,  it is increasingly evident that such techniques, while useful, will face fundamental stumbling blocks in providing scalability and responsiveness.

We propose to take a fundamentally different approach to tame this data deluge and deliver next-generation capabilities for scalable, accurate, and responsive decision making in a complex environment.  Our vision is to establish the LEarning and Algorithms for People and Systems (LEAPS) Center that will bring principled sublinear approaches to decision making in complex environments.

At the heart of our approach is the insight that there is an imminent need for scalable algorithmic methods that provide desirable performance speedups while sacrificing accuracy in a controlled fashion. Our “moonshot” goal is to develop cloud deployable, usable, and scalable sublinear systems that can enable diverse science and engineering applications to achieve orders-of-magnitude improved scalability, cost, and responsiveness compared to the status quo.