Una-May O'Reilly's Webpage

` Una-May O'Reilly's Webpage

Una-May O'Reilly

Principal Research Scientist
MIT Computer Science and Artificial Intelligence Lab

Menu

home
EVO-DesignOpt
teaching
contact info
biography
publications

I lead the Evolutionary Design and Optimization (EVO-DesignOpt) group.

If you are an MIT student, check EVO-DesignOpt pages for projects with possible UROP, UAP and M.Eng positions. New projects are being posted throughout Summer, Fall 2010.

EVO-DesignOpt conducts research in the areas of knowledge mining and experimental, simulation and human-in-the-loop optimization. Our group typically develops technically enhanced and specialized genetic algorithms in these contexts while incorporating machine learning, convex and linear optimization, regression and modeling techniques where complementary. Our research applies to problem domains of high complexity and real-world relevance such as automatic programming, distributed systems (eg. multicore), collaborative filtering, energy, and networks.

Projects in progress in 2010 include:

Knowledge mining Virtual Machine and Server data. A collaboration with VMWare.
Machine Learning to support an intelligent run time system. Collaboration with Prof. S. Amarasinghe of CSAIL, under DOE sponsorship of the ZettaBricks: A Language, Compiler and Runtime Environment for Anyscale Computing project.
Auto-Tuning Compilation for Multi-Core Systems
Wind Energy: Providing advanced intelligent information technology for wind farm assessment, operation and maintenance. This involves "mining" site-based data (including measurements from met towers, anemometers, control directives, status updates as diagnostics or direct sensor measurements) to reveal patterns, explanatory models, forecasts plus diagnostic and prognostic alerts.
Knowledge mining consumer satisfaction data: Developing innovative techniques that model people's preferences, find preference segments within a group of people and optimize designs or outcomes for specialized groups. This project's goals intersect with those of collaborative filtering because of the desire to find groups of people with similar tastes.
Optimization of Sparse Matrix Algebra Computation on High Performance Distributed Systems
Optimization of Distributed Sensor Network Fusion Strategies

History: I recently co-instructed Robotics: Science and Systems (6.141) at MIT in Spring 2010.

In 2006 I served as VP of Technology at Icosystem Corporation and led the Nymbler project.

You may know me by my past membership in the Humanoid Robotics Group.