Biographical Sketch

I am currently part of the Intelligent Data Understanding (IDU) Group, where my research is focused on three main project areas:

1. Supporting the ground-based diagnostics project for Ares I-X of the Constellation program, via development and deployment of data-driven technologies.
2. Supporting the IVHM project of the Aviation Safety Program for engine anomaly detection by the deployment of existing algorithms (IMS/Orca) and development of new ones, in order to provide an additional level of predictive and prognostic capability for anomaly detection.
3. Research & development in the area of optimal alarm systems, in which level-crossings of a univariate linear dynamic system can be predicted with the fewest false alarms possible. This research has application to the programs mentioned above, and potentially others.

I previously worked on the development of data-driven failure/fault detection algorithms for the liquid propulsion-based SSME (Space Shuttle Main Engine) for the ISHM project of the Exploration Technology Development Program (ETDP). I also helped with textual data mining to classify and identify unknown recurring anomalies from airline incident reports using existing and novel methods.

I was previously part of the Embedded Decision Systems Group where I worked on the prediction of operator intent and optimization of algorithms based upon Hidden Markov Models for telepresence of remotely supervised humanoid robots (Robonaut).

My other research interests include the intersection of the following areas:

1. Control theory
2. Mathematical Statistics (primarily level-crossing theory and optimal alarm systems)
3. Machine learning

Favorite Approaches

1. From Control Theory: Linear Gaussian Dynamic Systems and Stochastic State (Continuous/Discrete) Estimation, i.e. Hidden Markov Models, Kalman Filtering. Optimal Control Theory as related to Optimal Alarm-Based Control Systems and Optimal Reference Modification (potentially applicable to joint research with IRAC for mitigation and modified aircraft operation ).
2. From Mathematical Statistics: Optimal Alarm Systems (both research & development, please see my select pubs for more detail), Discretization of Continuous Time Level-Crossing Formulae for the Design of Optimal Alarm Systems.
3. From Machine Learning: Standard Bayesian and parametric approaches, probabilistic graphical modeling, system identification through dynamical learning, particle filtering.

Educational Background

Ph.D., Mechanical Engineering (2004)
University of California at Berkeley
Dissertation: Optimal Prediction, Alarm, and Control in Buildings Using Thermal Sensation Complaints

M.S., Mechanical Engineering (2000)
University of California at Berkeley
Thesis: Optimized Response to Thermal Sensation Complaints in Buildings

B.S., Mechanical Engineering (1992)
Carnegie-Mellon University