Program autotuning is the process of optimizing a program’s internal parameters to maximize or minimize a certain objective function, such as speed or accuracy. Currently, there is no systematic method for autotuning programs across multiple projects. OpenTuner, developed at CSAIL as a possible solution to this problem, is an open source framework for building domain-specific program autotuners. The goal of this project is to develop a system for OpenTuner to quickly and reliably evaluate a program’s search space representations. Given a set of candidate encodings, this project aims to build a tool that helps users determine the most optimal sequence of encodings to search over.

I spent this past summer at Google designing and building an evaluation system for content matching as part of the new App Indexing for Search platform. Prior to that, I have worked for various financial firms in New York City, using large historical data sets to create predictive models.