Email: < my first name > AT < my last name > DOT com
About Me
I am a Data Scientist working on applications of deep learning and Big Data. In the past few years I have worked on solving problems from industries such as social networks and advertising, network security, healthcare, finance, etc. using deep learning, machine learning and Big Data. Currently I am working at Twitter, and prior to that I worked at Cisco Systems' San Francisco Innovation Center (SFIC). I received my PhD in Operations Research from the University of Wisconsin-Madison in 2011. My PhD work was on theoretical and computational aspects of mathematical optimization, especially mixed integer linear and nonlinear programming problems. During my PhD I was also affiliated with Wisconsin Institute for Discovery (WID) and the French Institute for Research in Computer Science and Automation (Institut National de Recherche en Informatique et en Automatique, INRIA). Also I was a National Science Foundation (NFS) Grantee at the San Diego Supercomputer Center in 2007 and a Research Intern at IBM T.J. Watson Research Lab in 2008. After graduate school and before my position at Cisco I was a Scientist at Opera Solutions.
Work Experience
Senior Data Scientist, Twitter Cortex and CUAD
San Francisco, CA
June 2015 - Present
I am working on applications of deep learning and natural language processing (NLP). Some of my recent works for Twitter include: a deep neural architecture model for classification of abusive tweets; a Named Entity Recgonition (NER) systems for tweets based on word2vec word embeddings and bidirectional LSTMs; vectorizing users based on the Twitter's follow graph using skipgram; a language model for sequence of actions in user sessions based on Recurrent Neural Networks (RNNs) for sessions clustering and action prediction; a deep encoder-decoder based model for vectorizing tweets; and a video recommendation engine for promoted videos.
Senior Data Scientist, Cisco, San Francisco Innovation Center (SFIC)
San Francisco, CA
January 2013 - June 2015
I was a member of the Talos team of the Security Business Group. Using the Big Data technology and machine learning techniques I contributed to designing, building, and improving Cisco’s security appliances and technologies.
Scientist, Opera Solutions
San Diego, CA
July 2011 - January 2013
I performed applied research on a variety of predictive analytics problems coming from different industries such as healthcare and finance. I developed predictive signals and machine learning models for problems such as hospital readmission predictions for a major U.S. hospital chain and predicting the probability of credit default for a major U.S. bank.
Research Intern, IBM T.J. Watson Research Lab
Yorktown Heights, NY
May 2008 - September 2008
My research was on parallel branch and bound algorithms for mixed integer linear programming problems and also primal heuristics for these problems. I developed two novel primal heuristics for integer programming, namely Randomized Rounding and Pivot-and-Fix. These primal heuristcs have been in a part of COIN-Cbc since 2009.
NSF Grantee, San Diego Supercomputer Center
La Jolla, CA
June 2007 - August 2007
Under the NSF grant for Cyberinfrastructure Experience for Graduate Students (CIEG) I did research on high performace computing for mixed integer programming. I developed PMaP (Parallel Macro Partitioning) which is a parallel solver for mixed integer programs on shared-memory parallel computing frameworks.
Research Assistant, University of Wisconsin-Madison
Madison, WI
Sep 2006 - July 2011
As a PhD student, I performed theoretical and computational research on different aspects of mathematical optimization including mixed integer linear and nonlinear programming, parallel computing in mathematical programming, and global optimization. I also worked on the application of parallel computing in mixed integer linear programming. My PhD thesis was on theoretical and computational aspects of linear convexifications for multilinear function in optimization problems. I studied the strength of convex relaxations for nonconvex functions in general and multilinear functions in specific.
Education
PhD, Operations Research, University of Wisconsin-Madison, 2011. Advisor: Jeff Linderoth, Co-Advisor: Jim Luedtke
MSc, Operations Research, University of Wisconsin-Madison, 2008
MSc (Course Work), Information Technology, Amirkabir University of Technology (Tehran Polytechnic), 2006.
BSc, Industrial Engineering, Amirkabir University of Technology (Tehran Polytechinic), 2004.
Research Interests
Big Data and parallel computing: Big Data algorithms and storage, distributed graph databases and algorithms, Hadoop ecosystem and Map-Reduce, in-memory processing of Big Data, design of Big Data stack for organizations, parallel distributed and multi-threaded computing in optimization.
Machine learning: statistical machine learning, semi-supervised learning, application of graph theory in machine learning, post-processing, application of machine learning in network security, healthcare, and finance.
Mathematical programming and optimization: mixed integer linear and nonlinear programming theory, nonlinear optimization, large-scale optimization.
Selected Publications
James Luedtke, Mahdi Namazifar, and Jeff Linderoth. “Some Results on the Strength of Relaxations of Multilinear Functions”, Mathematical Programming, 136(2): 325-351, 2012. (pdf)
Pietro Belotti, Andrew Miller, Mahdi Namazifar, “Valid Inequalities and Convex Hulls for Multilinear Functions”, Electronic Notes in Discrete Mathematics 36: 805-812, 2010. (pdf)
Pietro Belotti, Andrew Miller, Mahdi Namazifar, “Linear inequalities for bounded products of variables”, SIAG/OPT Views-and-News 22(1), 1-7, 2011. (pdf)
Mahdi Namazifar and Andrew Miller. “A Parallel Macro Partitioning Framework for Solving Mixed Integer Programs”, Lecture Notes in Computer Science, 5015, 343-348, 2008. (pdf)
Mahdi Namazifar and Mohammad H. Taghavi, “Histogram Adjustment Using Mixed Integer Programming”, Proceedings of the 1st Technical and Analytical Conference at Opera Solutions, April 2012, San Diego, CA.
Pietro Belotti, Andrew Miller, and Mahdi Namazifar, “Valid Inequalities for Sets Defined by Multilinear Functions”, Proceedings of the European Workshop on Mixed Integer Nonlinear Programming (EWMINLP), April 2010, Marseille, France. (pdf)
Mahdi Namazifar, Robin Lougee, Andrew Miller, and John Forrest, “Randomized Rounding; A Primal Heuristic for General Mixed Integer Programming Problems”, Manuscript, August 2009. (Randomized Rounding is implemented in and is a part of COIN-Cbc) (pdf)
Mahdi Namazifar, Robin Lougee, John Forrest, “Pivot-and-Fix; A Mixed Integer Programming Primal Heuristic”, Manuscript, October 2009. (Pivot-and-Fix is implemented in and is a part of COIN-Cbc) (pdf)
Mahdi Namazifar, “Strong Relaxations and Computations for Multilinear Programming”, PhD Thesis, University of Wisconsin-Madison, 2011. (pdf)
Patents
Mahdi Namazifar, Wen Zhang, and Yan Zhang, “System and Method for Grouping Medical Codes for Clinical Predictive Analytics”, Patent Pending.
Mahdi Namazifar and Mohammad H. Taghavi, “Systems and Methods for Adjusting Distributions of Data Using Mixed Integer Programming”, Patent Pending.
Mahdi Namazifar, “Detecting Randomly Generated Strings; A Language Based Approach”, DEFCON23, Auguest 2015 (coming up), Las Vegas, NV. (Abstract, Slides, Video)
Mahdi Namazifar, “Distributed Graph-Based Entity Resolution Using Spark ”, Spark Summit East, March 2015, New York, NY. (Video, Slides)
Jeff Linderoth, Jim Luedtke, and Mahdi Namazifar, “Multi-term Relaxations for Multi-linear Programs”, Seminar Series, Priority Research Centre for Computer-Assisted Research Mathematics and its Applications, University of Newcastle, November 2012, Newcastle, Australia. (Invited)
Jeff Linderoth, Jim Luedtke, Ashutosh Mahajan, Mahdi Namazifar, “Solving mixed integer polynomial optimization problems with MINOTAUR”, 21st International Symposium on Mathematical Programming (ISMP), August 2012, Berlin, Germany. (Invited)
James Luedtke, Jeff Linderoth, and Mahdi Namazifar, “Strong Polyhedral Relaxations of Multilinear Functions”, 4th INFORMS Optimization Society Conference, February 2012, Coral Gables, FL. (Invited)
Jeff Linderoth, Jim Luedtke, and Mahdi Namazifar, “Practical Polyhedral Relaxations for Multilinear Programs”, 16th Combinatorial Optimization Workshop, January 2012, Aussois, France.
James Luedtke, Mahdi Namazifar, and Jeff Linderoth, “Strong Linear Relaxations for Global Optimization Problems with Multilinear Terms”, Department of Energy Applied Mathematics Program Meeting, October 2011, Washington, DC.
James Luedtke, Jeff Linderoth, and Mahdi Namazifar, “Strong Polyhedral Relaxations for Multilinear Programs”, SIAM Conference on Optimization, May 2011, Darmstadt, Germany.
Mahdi Namazifar, Pietro Belotti, and Andrew Miller, “Convex Outer Approximations for Bounded Multilinear Functions”, 10th INFORMS Annual Meeting, November 2010, Austin, TX.
Mahdi Namazifar, James Luedtke, and Jeff Linderoth, “New Linear Relaxations for Quadratically Constrained Quadratic Programming Problems”, 10th INFORMS Annual Meeting, November 2010, Austin, TX.
James Luedtke, Jeff Linderoth, and Mahdi Namazifar, “Comparisons of the Relative Strength of Relaxations for Bilinear Functions in Global Optimization”, 10th INFORMS Annual Meeting, November 2010, Austin, TX.
Pietro Belotti, Andrew Miller, and Mahdi Namazifar, “Valid inequalities for sets deﬁned by multilinear functions”, Toulouse Conference on Global Optimization (TOGO), September 2010, Toulouse, France. (Plenary)
Andrew Miller, Pietro Belotti, and Mahdi Namazifar, “Strong Formulation for Multi-Linear Sets”, 7th Mixed Integer Programming Workshop, July 2010, Atlanta, GA. (Invited)
James Luedtke, Jeff Linderoth, and Mahdi Namazifar, “Relaxations of Multilinear Functions in Mixed-Integer Nonlinear Programming”, 7th Mixed Integer Programming Workshop, July 2010, Atlanta, GA. (Invited)
Pietro Belotti, Andrew Miller, and Mahdi Namazifar, “Valid Inequalities for Sets Deﬁned by Multilinear Functions”, European Workshop on Mixed Integer Nonlinear Programming (EWMINLP), April 2010, Marseille, France. (Invited)
Andrew J. Miller, Mahdi Namazifar, and Pietro Belotti, “Valid Inequalities and Convex Hulls for Multilinear Functions”, International Symposium on Combinatorial Optimization (ISCO), March 2010, Hammamet, Tunisia.
Pietro Belotti, Andrew J. Miller, and Mahdi Namazifar, “Valid Inequalities for Bounded Multilinear Functions”, Workshop on Integer Programming and Combinatorial Optimization, January 2010, Aussois, France.
Jeff Linderoth, James Luedtke, Mahdi Namazifar, and Andrew Miller, “Tighter Relaxations for Global Optimization Problems with Multilinear Terms”, 9th INFORMS Annual Meeting, October 2009, San Diego, CA.
Mahdi Namazifar, Pietro Belotti, and Andrew Miller, “Convex Envelopes for Bounded Multilinear Functions”, 9th INFORMS Annual Meeting, October 2009, San Diego, CA.
Mahdi Namazifar, Robin Lougee-Heimer, and John Forrest, “Pivot-and-Fix; A New Primal Heuristic for Mixed Integer Programming”, 20th International Symposium on Mathematical Programming (ISMP), August 2009, Chicago, IL. (Invited)
Jeff Linderoth, Mahdi Namazifar, and James Luedtke, “Strong Relaxations and Computations for Global Optimization Problems with Multilinear Terms”, 20th International Symposium on Mathematical Programming (ISMP), August 2009, Chicago, IL. (Invited)
Andrew Miller, Pietro Belotti, and Mahdi Namazifar, “Valid Inequalities, Separation, and Convex Hull for Multilinear Functions”, 20th International Symposium on Mathematical Programming (ISMP), August 2009, Chicago, IL. (Invited)
Mahdi Namazifar, Pietro Belotti, and Andrew Miller, “Linear Envelopes of Bounded Multilinear Function”, 6th International Conference on Integration of AI and OR Techniques in Constraint Programming for Combinatorial Optimization Problems, May 2009, Pittsburgh, PA. (Invited)
James Luedtke, Mahdi Namazifar, and Jeff Linderoth, “Generating Improved Relaxations for Global Optimization with Multilinear Terms”, Workshop on Computational Issues in Mixed Integer Nonlinear Programming (CIMINLP), March 2009, Marseille, France. (Invited)
Mahdi Namazifar, Robin Lougee-Heimer, and Andrew Miller, “Randomized Rounding; a Primal Heuristic for Mixed Integer Programming”, 11th INFORMS Computing Society Conference, January 2009, Charleston, SC.
Andrew Miller and Mahdi Namazifar, “A Parallel Macro Partitioning Framework for Mixed Integer Programming”, 11th INFORMS Computing Society Conference, January 2009, Charleston, SC.
Andrew Miller and Mahdi Namazifar, “A Parallel Macro Partitioning Framework for Solving Mixed Integer Programs”, Automatic Reformulation Search Project Workshop (ARS08), October 2008, Ecole Polytechnique, Paris, France.
Andrew Miller and Mahdi Namazifar, “A Parallel Macro Partitioning (PMaP) Framework for Large Mixed Integer Programs”, 8th INFORMS Annual Meeting, October 2008, Washington D.C.
Mahdi Namazifar and Andrew Miller, “Randomization and Rounding Heuristics for Finding Feasible Solutions of General Integer Programs”, 8th INFORMS Annual Meeting, October 2008, Washington D.C.
Andrew Miller and Mahdi Namazifar, “A Parallel Macro Partitioning Framework for Solving Mixed Integer Programs”, 5th International Conference on Integration of AI and OR Techniques in Constraint Programming for Combinatorial Optimization Problems, May 2008, Paris, France.
Mahdi Namazifar and Andrew Miller, “A Parallel Macro Partitioning Framework for Solving Mixed Integer Programming Problems”, 2nd INFORMS Optimization Society Conference, March 2008, Atlanta, GA.
Andrew Miller and Mahdi Namazifar, “Parallel Local Search Methods for Mixed Integer Programs”, 7th INFORMS Annual Meeting, November 2007, Seattle, WA.
Mahdi Namazifar, James Luedtke, Jeff Linderoth, “Strength of Relaxations of Multilinear Functions for Mixed-Integer Nonlinear Programming”, 7th Mixed Integer Programming Workshop, July 2010, Atlanta, GA.
Mahdi Namazifar, James Luedtke, and Jeff Linderoth, “Improved Relaxations for Mixed Integer Nonlinear Programming Problems with Multilinear Terms”, 6th Mixed Integer Programming Workshop, June 2009, Berkeley, CA.
Mahdi Namazifar, “Towards a Parallel Macro Partitioning Framework for MINLP Problem”, IMA Workshop on Mixed Integer Nonlinear Optimization: Algorithmic Advances and Applications, November 2008, Minneapolis, MN.
Mahdi Namazifar and Andrew Miller, “High-Level Branching Methods for Solving Mixed Integer Programs in Parallel”, NSF CMMI Engineering Research and Innovation Conference, January 2008, Knoxville, TN.