Carnegie Mellon University, The Robotics Institute
Ph. D. Robotics, 1997
University of California, Berkeley
Dept. of Electrical Engineering and Computer Science, 1989.
Win-Vector, LLC. San Francisco, California
Principal Consultant, 2008-Present
Win-Vector conducts data analysis and statistical research for a variety of private sector clients, particularly in the biotech, finance, and internet sectors. Engagements that I have led or been involved in have included: revenue attribution for Google ad-words, customer modeling from online transaction data, product recommendation systems, and loan risk modeling.
I am the co-author of Practical Data Science with R, a introductory data science text written from a practitioner's perspective.
I have developed original training content for a number of popular training courses including:
- DataCamp’s Supervised Learning in R: Regression course
- Udemy’s Introduction to Data Science course
- EMC Corporation’s popular Data Science and Big Data Analytics course
- Proprietary Data Science training courses for a number of large corporations
Contributor to CRAN R projects:
vtreat: a data treatment package for predictive modeling
WVPlots: a package of common
ggplot2based plots for data science
Occasional contributor to the Win-Vector blog; started the popular Statistics to English Translation series.
Quimba Software, Inc. San Francisco, California
Co-founder, Owner. 2001-2008
Principal Investigator, Command Improvisation Support System for Disaster Managers
Phase I STTR, National Science Foundation.
In collaboration with Dr. Larry Beutler and Zeno Franco, The Pacific Graduate School of Psychology.
In a truly catastrophic situation, existing disaster response plans may prove to be inadequate, impossible to execute, or an appropriate plan may simply not exist. In such situations, it is necessary for disaster responders to improvise -- that is, to deviate from standard procedure in order to accomplish necessary goals. The focus of this Phase I STTR is to investigate the design of a Command Improvisational Support System (CISS) that will aid an EOC-level disaster manager in recognizing the situational cues that indicate that deviation from standard procedure, or a return to standard procedure, is indicated, and to assist in communicating the necessary changes to disaster response team.
Senior Investigator, Automated Data Staring across Functional Boundaries
SBIR Phase I and Phase II, US Air Force Research Laboratory, Rome, NY.
The goal of this effort is to build an information sharing infrastructure, based on Web Services, specifically intended for deployment across diverse functional boundaries. This system is being developed in the Bio-surveillance domain, where it is envisioned that participants would potentially include Federal, Military, State and Local emergency responders.
I was responsible for developing the software component to automatically generate and configure the operational Web Services that are used to instantiate the data sharing chain, In particular I am responsible for developing the knowledge representations that will be used to automatically generate web services for enforcing the diverse security and other data sharing constraints that are imposed on the various organizations that participate in the data-sharing chain.
Principal Investigator, Source Characterization for Information Fusion
BAA contract , US Air Force Research Laboratory, Rome, NY
SBIR Phase I, US Air Force Research Laboratory, Rome, NY
The goal of this twenty seven month research effort was to discover and implement user-based notions of relevance that can be used for more efficiently identifying information that is pertinent to a user's goals and needs. Originally, our research effort focused on automated methods for characterizing the relevance of an information source to a user's needs. The effort evolved into the foundations of an Online Research Assistant that can
- Disambiguate ambiguous query terms
- Retrieve search results that match a user's specific criteria, both objective and subjective
- Learn the most effective sources for a user's ongoing information needs.
To demonstrate our results, we developed a prototype advisable system for online research, based on the Buddy metasearch engine, which was developed at AFRL. This technology was specifically developed to help intelligence analysts manage their ongoing information needs.
Optivo, Palo Alto, California
Staff Software Engineer, Numerics, 2000-2001
Developed the core numerics engine for Optivo's Live Price Testing and Price Optimization product. The Optivo Price Manager monitors market demand for products on online channels in real or semi-real time, and adjusts prices accordingly to optimize the merchant's business objectives. A trial version of this system was successfully deployed for several online merchants.
Designed the testing and optimization algorithms; the testing algorithms balance issues of statistical significance and test-cost, as well as compensating for non-stationarity in the pricing environment. Wrote the Java-based optimization module, and wrote or worked on several SQL and PL/SQL scripts and procedures in support of the optimization module. Worked closely with the User Interface team in designing the online Analytics reports which a merchandiser would use to monitor the performance of his or her products while using our system. Worked with the Product Management and UI teams on issues of the accessibility and ease of use of our system by non-technical customers, such as merchandisers.
Artificial Intelligence Center, SRI International, Menlo Park, California
Computer Scientist , 1997 through 2000
Automated Capture of Design Rationale
Took over the development of a C based monitoring system for capturing the actions of the users of a commercial CAD tool. Wrote a Lisp inference engine for abstracting, recording, and interpreting the actions of design engineers captured by the monitoring system. The resulting system produces a history of the design process, registered to the CAD design, which is annotated with interpretations of the designer's actions. SRI has a patent pending on this system.
Terrain-Based Evidential Reasoning Techniques for Tracking
Developed a system for incorporating knowledge of terrain characteristics into an existing Bayesian tracking algorithm. The resulting system produces fewer false alarms than a similar system that does not use terrain characteristics.
Similarity Based Document Indexing and Retrieval
Developed a similarity-based indexing scheme for organizing and retrieving documents in a knowledge base of intelligence analysts' reports.
Design of Parts Orienting Devices
Dissertation Research, CMU
Designed, analyzed and tested a prototype sensorless parts orienting device for use in robotic assembly lines, and developed associated algorithms which reduced the search for plans from extensive numerical computation to graph search and linear programming. The resulting reorientation plans are robust to uncertainties in the part's initial state and in the coefficient of friction, as well as to small inaccuracies in manipulator calibration.
Characterization of shape invariant material properties
With Eric Krotkov, The Robotics Institute, CMU,
and Roberta Klatzky, Department of Psychology, CMU.
We developed and implemented an algorithm for acoustically describing the material properties of an object based on spectrograms of impulsive contact with object. This resulted in a method of compactly storing data to synthesize physically plausible sounds of tapped objects for virtual reality applications.
Simulation of Robot Juggling
With Michael Erdmann, The Robotics Institute and School of Computer Science, CMU.
Modeled and simulated impact dynamics of polygonal objects for incorporating into simulations of robot juggling. Implemented the dynamics module of an animated simulation.
Nina B. Zumel and Vladimir Gorelik. Method and Apparatus for Dynamic Pricing during Inventory Reduction.
Nina B. Zumel and Vladimir Gorelik. Method and Apparatus for Automated Demand Trend Correction in Dynamic Pricing.
Vlad Gorelik, Andrew Atherton, and Nina Zumel. Method and Apparatus for Automatic Pricing in Electronic Commerce.
Vlad Gorelik, Andrew Atherton, and Nina Zumel. Method and Apparatus for Determining Customer Lifetime Value and Setting Price.
Karen L. Myers, Nina B. Zumel, and Pablo Garcia. Automated Capture of Rationale for Detailed Design Process.
John Mount and Nina Zumel. The vtreat R package: a statistically sound data processor for predictive modeling. The Journal of Open Source Software, 3(23), 584. (2018)
Zeno E. Franco, Nina Zumel, Kathy Blau, Knute Ayhens-Johnson, and Larry Beutler. Causality, covariates and consensus in ISCRAM research: towards a more robust study design in a transdisciplinary community. International Journal of Emergency Management, Vol. 5, pp 100-122. (2008)
Nina Zumel, Zeno Franco and Larry Beutler. Improvisation as a Training Framework for Emergency Managers. Presented at the 13th International Command and Control Research and Technology Symposium (ICCRTS 2008), Bellevue, WA June 17-19, 2008.
Zeno Franco, Nina Zumel and Larry Beutler. A Ghost in the System: Integrating Validity and Methodology Considerations from Behavioral Sciences into Disaster Research. In B. Van de Walle, P. Burghardt, & C. Nieuwenhuis (Eds.), Proceedings of the 4th International Conference on Information Systems for Crisis Response and Management (ISCRAM2007) (pp. 115-124). Brussels: VUBPRESS.
Alessandro Saffiotti, Nina B. Zumel, and Enrique H. Ruspini. Multi-Robot Team Coordination using Desirabilities. 6th International Conference on Intelligent Autonomous Systems (IAS-6), Venice, Italy, July 25-27, 2000.
Karen L. Myers, Nina B. Zumel, and Pablo Garcia. Acquiring Design Rationale Automatically. Artificial Intelligence for Engineering Design, Analysis, and Manufacturing (2000) vol. 14, pp. 115-135.
Karen L. Myers, Nina B. Zumel, and Pablo Garcia. Automated Capture of Rationale for the Detailed Design Process. Conference on Innovative Applications of Artificial Intelligence, July 1999.
N. B. Zumel and M. A. Erdmann. Nonprehensile Manipulation for Orienting Parts in the Plane. 1997 IEEE International Conference on Robotics and Automation, April 1997.
N. B. Zumel and M. A. Erdmann. Nonprehensile Two Palm Manipulation with Non-Equilibrium Transitions between Stable States. 1996 IEEE International Conference on Robotics and Automation.
E. Krotkov, R. Klatzky, and N. Zumel. Analysis and synthesis of the sounds of impact based on shape invariant properties of materials. 13th International Conference on Pattern Recognition, 1996.
E. Krotkov, R. Klatzky, and N. Zumel. Robotic Perception of Material: Experiments with Shape-Invariant Acoustic Measures of Material Type. Experimental Robotics IV, O. Khatib and K. Salisbury, editors. Lecture Notes in Control and Information Sciences, 200. Springer-Verlag, 1996. Proceedings of Conference, June-July 1995.
N. B. Zumel and M. A. Erdmann. Balancing of a Planar Bouncing Object. 1994 IEEE International Conference on Robotics and Automation.