FEDOR DZEGILENKO, Ph.D.
**** ********** **** ***, *** Jose, CA 95123; e-mail: abg2v0@r.postjobfree.com
ph: 408-***-**** (mobile).
OBJECTIVE:
( Seeking a challenging position where 10+ years of experience in
development of advanced mathematical &
statistical predictive algorithms will help to solve critical
business problems.
SUMMARY OF QUALIFICATIONS:
( Developed real-time precision targeting statistical algorithms to
identify potential "defectors" for the TurboTax
online product. These algorithms significantly increased conversion
rates. Estimated additional revenue is ~
50 million.
( Designed and developed a decision support system for eBay sellers
which significantly increased eBay's and
seller's profit from the online auction. The system was based on a
combination of statistical modeling and
optimization techniques.
( Performed statistical modeling of click-stream patterns at Intuit's
Turbo Tax site. Suggested ways to leverage
the modeling results to improve this site's features.
( Developed successful algorithms for the short-term prediction of
time-series with financial data and applied it
to Forex trading.
( Used data mining/machine learning/econometric techniques to develop
optimal pricing strategies for retailers,
manufacturers, and Internet Service Providers.
. Highly proficient in various mathematical & statistical modeling
techniques such as Time Series Analysis & Forecasting, Nonlinear
Optimization, Monte-Carlo Simulations, Markov Processes, Kalman Filter,
Neural Nets, Pattern Recognition, Clustering & Classification, Decision
Trees, Logistic Regression, Multivariate Statistics, and Bayesian
Techniques.
. Published/presented more than 40 papers at major international
journals/symposiums.
. Author of several patents in the fields of behavioral internet-based real-
time targeting, online auctions, price optimization, and Nanotechnology.
. R,SAS, Java, C, Matlab, GAMS, Weka
WORK EXPERIENCE:
Intuit, Mountain View, CA
01/2008-
present
Principal Statistician (Business Intelligence and Optimization):
Defined the overall statistics strategy for a portfolio of very different
businesses within Intuit.
( Designed and developed statistical algorithms to identify potential
"defectors" for the TurboTax online product using click-stream data. This
allowed for real-time precision targeting of potential defectors so that
assistance would be offered to them. That capability resulted in increased
conversion rates and additional revenue.
( Developed a set of statistical models to identify up-sell and cross-sell
opportunities within QuickBooks products.
These models were deployed in marketing campaign and generated a
significant lift in sales.
( Developed a probabilistic model of user's browsing behavior at Intuit's
Turbo Tax site. The knowledge of the patterns of user's behavior made it
possible to optimize the flow of information between different sections of
the site and to improve user's experience.
( Developed Bayesian attribute-based approach to match small business
offers to customers. The algorithm provides optimal recommendations to
mobile phone users within Intuit's network.
eBay, Campbell, CA
03/2006-
01/2008
Principal Research Scientist (eBay's Research Lab):
Responsible for the development of advanced mathematical algorithms that
improve eBay's profitability and reduce fraud risks.
( Developed a statistical methodology to predict the maximal bid on an
online auction using real-life data. Based on that model, developed and
implemented an optimization procedure to maximize eBay's and seller's
profits from the auction. This combination of the econometric model and the
optimization resulted in a decision support system which provided guidance
to the seller at the time of initial listing. This significantly increased
eBay's profits.
The expected extra gain in profits was more than 50% per auction.
( Developed an algorithm for fraud-detection which dynamically updated the
probability of fraud with every new transaction. This novel approach allows
for fraud detection in real time as opposed to standard "static"
algorithms.
Tyche Capital Management, San Jose, CA
07/2005- 05/2007
Founder, Managing Partner:
( Developed short-term trading models using a combination of statistical,
machine learning, and optimization techniques. Successfully applied these
models to trade in currency exchange markets.
True Demand Software, Los Gatos, CA
07/2005 -03/2006
Chief Scientist:
Responsible for the development and management of statistical predictive
modeling solutions when applied to supply chain forecasting and planning.
( Developed a statistical methodology to predict the probability of out-
of-stocks occurrences which leveraged real-time point-of-sales and
inventory data.
( Developed econometric models to predict demand uncertainty for
retailers and manufacturers.
( Proposed solutions to estimate future demand for newly introduced
products.
DemandTec, San Carlos, CA
03/2002 -07/2005
Senior Scientist:
Developed econometric models which described consumer demand for retailers.
Created algorithms for price optimization. Performed intelligent customer
and product segmentation. Managed multiple science projects for the
Engineering, Professional Services, and Sales departments. These
responsibilities ranged from designing technical specs to final
testing/validation. Assessed risks of new product features from technical
and business standpoints.
( Developed a short- product-life-cycle econometric model for the
markdown process and used a dynamic
programming based approach to find the optimal markdown policy.
Applied Kalman filter methodology to
update forecast.
( Significantly increased predictive abilities of the existing
econometric model of consumer demand by using
nonlinear multivariate regression via Bayesian shrinkage methodology
( Designed and implemented algorithms for clustering and classification
of products and their attributes. Created
algorithms for customer segmentation using several machine learning
techniques (Decision Trees, Support
Vector Machines, Neural Nets, Logistic Regression, Hierarchical
Clustering).
( Developed algorithms for market basket analysis to identify cross and
up-sell opportunities.
( Applied Neural Networks to predict future sales.
( Developed algorithms to account for missing observations using
Bayesian techniques.
( Performed optimization of product prices using Nonlinear and Mixed
Integer techniques.
( Researched and recommended new methodologies aimed at enhancing
predictive accuracy of existing models
and product's stability and scalability.
( Analyzed feedback from customers and proposed quantitative tests and
strategies to enhance sales.
XACCT Technologies, Santa Clara, CA
12/2000-11/2001
Senior Member of R&D Department, Technical Staff:
Researched and recommended various business application, using high
frequency data for internet protocols (IP).
( Analyzed various IP metrics collected from computer networks using
machine learning and data mining
techniques (Logistic Regression, Decision Trees and Clustering).
( Proposed and implemented multiple business applications such as
differential billing, churn/retention analysis,
fraud detection/prediction, optimal pricing models, and web
analytics.
( Worked closely with customers to explain technical aspects of the
methodology.
ePLANNING, San Jose, CA
03/2000-12/2000
Senior Member of a Science Group:
Responsible for the development of quantitative portfolio optimization
tools for financial planners.
( Designed and implemented a package for financial portfolio simulation
(Monte-Carlo based) and asset
allocation (quadratic optimization) using Value-at-Risk methodology.
( Used various Stochastic Volatility models (GARCH, ARCH, hidden Markov
models) to account for "fat tails"
in distribution of stock returns.
( Created short-term stock prediction strategies using pattern
recognition and nonlinear time-series analysis.
( Modeled option prices by performing Monte-Carlo simulation of
underlying assets.
( Presented results to prospective buyers of the product.
NASA Ames Research Center/MRJ Technology Solutions, Moffett Field, CA
09/1997-03/2000
Research Scientist, Information Technology Modeling and Simulation Group:
Performed fundamental research in the areas of nonlinear dynamics and
Nanotechnology:
( Developed and coded a successful algorithm for time-series prediction
based on the Chaos Theory.
Applied it to predict the level of solar activity and to perform
nonlinear analysis of high frequency
financial time series.
( Used High Performance Fortran to parallelize solutions for the multi-
dimensional Schrodinger equation on
Cray.
( Utilized neural nets to model potential energy surfaces.
( Demonstrated the feasibility of a new class of carbon-based materials
for space applications using stochastic
molecular dynamics simulations.
Emory University, Atlanta, GA
11/1995-9/1997
Postdoctoral Researcher in Computational & Theoretical chemistry:
Carried out fundamental research in the area of Computational Quantum
Chemistry.
( Applied the random matrix theory to calculate rates of critical
chemical reactions.
( Worked on the development of a novel numerical method to solve
partial differential equations.
The Institute of Chemical Physics of the USSR Academy of Sciences, Moscow,
Russia 6/1988 -6/1991
Research Scientist:
( Conducted computational research in the area of nonlinear physics.
EDUCATION:
Department of Physics, The Chemical Physics Program at The Ohio State
University, Columbus, OH
( Ph.D. (Computational Chemical Physics), October 1995.
This Ph.D. thesis was chosen as the best example of how to correctly write
a Ph.D. Thesis for the Ohio State University Graduate Students Guide.
Department of Chemistry, Moscow State University, Moscow, Russia
( M.Sc. (Theoretical & Computational Chemistry), June 1988.
AWARDS:
( Spotlight achievement cash awards: eBay 10/2007, Intuit 04/2008,
04/2009, 07/2009, 05/2010.
( NASA Ames Award for Excellence in Research and Outstanding
Contributions to the Emerging Field of
Nanotechnology, March 2000.
( Ames Contractor Council Excellence Award from the NASA Ames Research
Center, October 1998.
PERSONAL:
( US citizenship, Chess Master, avid hiker and ping-pong player.