Computer Science Design Engineer
Resume:
*** ******* ** *** * *******@***.***
Cambridge MA, 02139 http://www.csail.mit.edu/ adrianc
Adrian Corduneanu
Education
Massachusetts Institute of Technology
Ph.D. candidate in Computer Science
Thesis supervisor: Prof. Tommi Jaakkola
Minor: Business Administration at Sloan School of Management
Massachusetts Institute of Technology
M.Sc., January 2002
Thesis: Stable Mixing of Complete and Incomplete Information
Thesis supervisor: Prof. Tommi Jaakkola
University of Toronto
B.Sc. in Computer Science and Mathematics, May 1999
With High Distinction, Dean s List Scholar
Fellowships and Awards
Awards
Best Paper Award, 19th Conference on Uncertainty in Arti cial Intelligence
2003
1999 Word Finalist, ACM Programming Competition, Eindhoven, The Netherlands
Outstanding Winner, 13th Mathematical Contest in Modeling
1997
Gold Medal, 37th International Mathematical Olympiad, Bombay, India
1996
Silver Medal, 36th International Mathematical Olympiad, Toronto, Canada
1995
Silver Medal, 10th Balkan Mathematical Olympiad, Nicosia, Cyprus
1993
Fellowships
1999, 2002 NSERC Canadian Fellowship for Graduate Research
2000 Presidential Fellowship, Massachusetts Institute of Technology
1996 1999 National Scholarship, University of Toronto
1996 1999 George Roderick Fraser Scholarship in Mathematics, University of Toronto
1997,1998 Galois Award in Mathematics, University of Toronto
1997, 1998 University of Toronto Scholar, University of Toronto
1998 Samuel Beatty Award in Computer Science and Mathematics, Univ. of Toronto
1998 Dr. James A. & Connie P. Dickson Scholarship in the Sciences and Mathematics,
University of Toronto
Tes Mossman Admission Scholarship, University of Toronto
1996
Research and Industrial Positions
Summer 2004 Research Intern, Microsoft Research Redmond, WA
Advisor: Dr. John Platt
Designed a novel robust real-time algorithm for super-resolution of images that en-
hances the resolution of text while being robust on other type of detail. Applications
include scalable UI s, real-time zooming of web interfaces, enhancing low-resolution
camera images of documents, or web graphics for printing.
1
Research and Industrial Positions (contd.)
Summer 2003 Research Intern, Microsoft Research Redmond, WA
Advisors: Dr. Hagai Attias and Dr. Eric Brill
Developed an entertaining conversational agent that extracts correct and coherent
English replies from the Internet in real-time, trained with machine learning techniques
to recognize dialog. The demo was highly regarded.
9/2002 5/2003 Visiting Researcher, University of Toronto Toronto, Canada
Advisor: Prof. Brendan Frey
Collaborated with a biology lab to deliver statistical algorithms for denoising microar-
ray data.
Summer 2001 Research Intern, Microsoft Research Redmond, WA
Advisor: Dr. Chris Meek
Implemented a discriminative algorithm for classi cation with Bayesian networks with
application to handwritten character recognition of Japanese.
Summer 2000 Research Intern, Microsoft Research Cambridge, UK
Advisor: Dr. Christopher Bishop
Developed and published a Variational Bayes algorithm for automatically determining
the number of components in a mixture of Gaussians model.
Summer 1999 Software Design Engineer, Microsoft Corporation Redmond, WA
Designed and developed a statistical code-optimization tool that reorders C++ struc-
tures to minimize the number of cache misses. The tool, trained on the memory access
pattern from sample runs, scored 1% benchmarked speed improvement on MS SQL
Server (US patent #6678805).
Summer 1998 Visiting Scientist, Johns Hopkins University Baltimore, MD
Advisor: Prof. Vassilis Digalakis
Developed an algorithm for adaptation of a speech recognition system to new speakers
from a very short sample of speech that increased speech recognition accuracy by 1%.
Other Research
9/2003 5/2004 Research Assistant, MIT Cambridge, MA
Advisor: Prof. Tommi Jaakkola
Pioneered a new class of algorithms for classi cation with labeled and unlabeled data
based on the information regularization principle.
Designed and implemented a Hidden Markov Model that produces a compact repre-
sentation of classes of genomic binding motifs.
9/2001 8/2002 Research Assistant, MIT Cambridge, MA
Advisor: Prof. Tommi Jaakkola
Introduced homotopy continuation methods to the machine learning community, with
application to learning from labeled and unlabeled data.
9/1998 6/1999 Research Collaboration, Johns Hopkins University University of Toronto
Advisors: Prof. F. Jelinek, Prof. S. Khudampur, and Prof. G. Hirst
Implemented and published a large-vocabulary language model that predicts the next
word from long histories clustered with a decision tree (10, 000 lines of C++ code)
2
Teaching Experience
Fall 2004 Recitation Instructor, Massachusetts Institute of Technology Cambridge, MA
Course: Machine Learning ( rst-year graduate core class)
Taught supplementary material in recitations, organized group grading sessions, su-
pervised students on their nal project.
Spring 1999 Recitation Instructor, University of Toronto Toronto, Canada
Course: Software Tools and Systems Programming (sophomore core class)
Conducted recitations, supervised laboratory sessions and examinations.
9/1996 5/1999 Recitation Instructor, University of Toronto Toronto, Canada
(6 semesters) Course: Calculus (freshman core class)
Conducted recitations and individual tutoring sessions, administered weekly quizes.
Publications
Refereed Conference Papers
Corduneanu, A. and Platt, J. C. (2005)
Learning Spatially-Variable Filters for Super-Resolution of Text
Submitted to the IEEE International Conference on Image Processing 2005
Corduneanu, A. and Jaakkola, T. (2004)
Distributed Information Regularization on Graphs
In Neural Information Processing Systems 2004
Corduneanu, A. and Jaakkola, T. (2003)
On Information Regularization
In Uncertainty in Arti cal Intelligence 2003, Best Paper Award
Corduneanu, A. and Jaakkola, T. (2002)
Continuation Methods for Mixing Heterogeneous Sources
In Uncertainty in Arti cal Intelligence 2002
Corduneanu, A. and Bishop, C. (2001)
Variational Bayesian Model Selection for Mixture Distributions
AI and Statistics 2001
Corduneanu, A. (1999)
A Pylonic Decision-Tree Language Model with Optimal Question Selection
Proceedings of the 37th Annual Meeting, Association for Computational Linguistics
Bocchieri, E., Digalakis, V., Corduneanu, A., Boulis, C. (1999)
Correlation Modeling of MLLR Transform Biases for Rapid HMM Adaptation
to New Speakers
Proceedings of ICAASP 1999
Digalakis, V., Corduneanu, A., et al. (1999)
Rapid Speech Recognizer Adaptation to New Speakers
Proceedings of the ICAASP 1999
Journal Articles
Corduneanu, A., Hsia, C., and O Donnell, R. (1997)
A Greedy Algorithm for Solving Meeting Mixing Problems
The UMAP Journal
Master Thesis
Corduneanu, A. (2001) Stable Mixing of Complete and Incomplete Information.
3
Technical Reports
Zatloukal, K., Corduneanu, A., Ladner, R., Grover, V., Meacham, S. (1999)
Improving Cache Performance by Structure Reordering
Technical Report from 1999 Microsoft internship
Peer Reviewing
IEEE Transactions on Pattern Analysis and Machine Intelligence (PAMI)
International Workshop on Arti cial Intelligence and Statistics
Professional Presentations
10/2004 PHZ Capital Partners Wayland, MA
Compact Characterization of Binding Sites via Hidden Markov Models
8/2004 Microsoft Research Redmond, WA
Adaptive Filtering for Real-Time Superresolution of Text
4/2004 Stochastic Systems Group Massachusetts Institute of Technology
Large-scale Information Regularization and Extensions
8/2003 Microsoft Research Redmond, WA
Internet-Powered Chat Agent
19th Conference on Uncertainty in Arti cial Intelligence
8/2003 Acapulco, Mexico
On Information Regularization
10/2002 Stochastic Systems Group Massachusetts Institute of Technology
Continuation Methods for Mixing Heterogeneous Sources of Information
11/2002 Probability and Statistical Inference Group University of Toronto
Continuation Methods for Mixing Heterogeneous Sources of Information
18th Conference on Uncertainty in Arti cial Intelligence
8/2002 Edmonton, Canada
Continuation Methods for Mixing Heterogeneous Sources
8th International Workshop on Arti cial Intelligence and Statistics
1/2001 Key West
Variational Bayesian Model Selection for Mixture Distributions
8/2000 Microsoft Research UK Cambridge, UK
Discrete Model Selection With Continuous Hyperparameters
37th Annual Meeting of the Association for Computational Linguistics
7/1999
A Decision-Tree Language Model
8/1998 The Center for Language and Speech Processing Johns Hopkins University
Correlation Modeling for Bias Adaptation
4/1998 Mathematical Association of America Regional Meeting Toronto, Canada
1997 Mathematical Contest in Modeling: A Meeting Mixing Problem
Programming Skills
Expert C++, Perl, Matlab, C#, Java, L TEX, Linux and Windows programming
A
References
Available upon request
4
Contact this candidate