stanford university, stanford, ca
m.s. degree in engineering economic systems and operations research in june XX.
ph.d. degree in management science and engineering june XX.
dissertation title: "multi-agent learning and coordination algorithms for distributed dynamic resource allocation."
dissertation advisor: nicholas bambos
massachusetts institute of technology, cambridge, ma
b.s. degree in mathematics in june 1997.
m.s. degree in systems science and control engineering from the department of electrical engineering and computer science in june 1998. masters thesis topic: context-sensitive planning for autonomous vehicles operating in complex, uncertain, and nonstationary environments.
experience
sun microsystems laboratories, menlo park, ca
april XX – present:
conceiving, developing and implementing self-managing and self-optimizing capabilities in computer systems, covering domains such as: cache-aware thread scheduling and cpu power management, dynamic sharing of cpu/memory/bandwidth, dynamic data migration in distributed storage systems, dynamic job scheduling and job pricing in cloud computing, dynamic user migration in distributed virtual environments, etc.
principal investigator for the adaptive optimization project since XX.
multiple patent applications filed, conference/journal papers published, multiple successful adaptive learning systems designed and implemented. the publicly available case studies are in the “technical reports” section of
intelligent inference systems corp., sunnyvale, ca research scientist
april XX – april XX: started a new research initiative in applying the acfrl algorithm and the previously developed multi-agent coordination algorithms to power control in wireless networks. published several conference papers on this topic. results demonstrate an improvement by more than a factor of 2 in comparison with the algorithms used in is-95 and cdmaXX standards.
april XX – april XX: wrote a phase i sttr proposal to the office of naval research and received funding for the topic of “perception-based co-evolutionary reinforcement learning for uav sensor allocation.” developed theoretical algorithms and designed a practical implementation strategy, which demonstrated excellent results in a high-fidelity robotic simulator. published a conference paper.
october 1998 – april XX: wrote a proposal to the nasa program in thinking systems and received multi-year funding for the topic of cooperation and coordination in multi-agent systems. developed, ｅｖａｌuated, and published new reinforcement learning algorithms for dynamic resource allocation among distributed agents operating jointly in complex, uncertain, and nonstationary environments.
fall XX: developed a new algorithm for single-agent learning in noisy dynamic environments with delayed rewards: actor-critic fuzzy reinforcement learning (acfrl). published a conference and a journal paper with a convergence proof for acfrl. us patent (number 6,917,925) was granted for the acfrl algorithm on july 12, XX.
chaincast inc., san jose, ca
aug XX – oct XX: conducted a survey of techniques for dynamic updating of multicasting trees and suggested a novel approach based on using multi-agent learning.
nasa ames research center, moffet field, ca summer 1998: designed a framework for multiple agents operating in a complex, uncertain, and nonstationary environment. agents learn to improve their policies using fuzzy reinforcement learning.
sri international, artificial intelligence center, menlo park, ca
summer 1998: developed a methodology for representing a replanning problem in the space of plans as a reinforcement learning problem.
bear, stearns & co., inc. - proprietory trading department, new york, ny
summer 1996, 1997: conducted a comprehensive study of time series forecasting models with neural networks. recommended a hybrid model combining best features of the existing models and implemented it in c++.
summer 1995: developed a stock forecasting system based on conventional econometric techniques and implemented it in sas language. gained exposure to various proprietary trading models.
alphatech, inc., burlington, ma
feb 1997 - may 1997: developed an algorithm for optimal control of macroeconomic systems described by simultaneous-time equations and implemented it in matlab.
arthur andersen, inc., boston, ma
feb 1996 - may 1996: developed an internal system dynamics cashflow model of startup businesses. gained experience in management level client interactions and in project presentation skills.
summer 1996: independently designed a game theoretic bid forecasting system in procurement auctions for a large construction company. the project involved extensive on-site client interactions during model development as well as a final presentation to the top level management.