Jeffrey Wong
Contact Phone: 650-***-****
Information E-mail: abo09b@r.postjobfree.com
EducationMS, Statistics, Jan 2011 - PresentBA, Economics, Sept 2007 - June 2011BS, Math &
Computational Science, Sept 2007 - June 2011Stanford University, Stanford, CA
Coursework includes: PhD series in Statistical Learning, PhD series in Applied Statis-
tics, Machine Learning, Parallel Programming, Time Series Analysis, Mathematical
Programming, Real Analysis, Applied Matrix Theory, Financial Mathematics
Portfolio http://stanford.edu/~jeffwong
Research Scientific Computing, Big Data Analysis, Game Theory, Algorithms and
Optimization,
Interests Graph Theory, Strategy and Tactics, Machine Learning, Statistical Learning
Programming General Languages: C, C++, C#, CUDA, Microsoft XNA, Java, LaTeX
Computational Languages: R, MATLAB, Python: numpy & scipy
Other: Hadoop, Cascading
Research Stanford University, Stanford, CA
Experience
Department of Electrical Engineering,
Civil & Environmental Engineering: Research AssistantJan 2011 to Present
Focus on data mining energy demand profiles and developing software infrastructure
for new research lab. Development of tsclust and fastVAR, R packages to handle time
series clustering and large scale VAR estimation. Developed a computing cluster for
research lab, and built software infrastructure to ease the analysis of a 80 GB dataset
for other team members.
Department of Mathematics: Research AssistantApril 2010 to Oct 2010
Using NVIDIA CUDA language to develop a parallel eigenvalue decomposition al-
gorithm for sparse matrices. Wrote a random matrix generator, parallel algorithm
for sparse matrix-vector multiplication, and implemented the Lanczos method for
eigenvalue decomposition on the GPU.
Work Apple:
Advanced Computational Group, Portland, OR
Experience
InternJune 2012 to Sept 2012
High performance computing in Apple's Advanced Computational Group under Dr.
Richard Crandall.
WeatherBill, San Francisco, CA
Math InternJuly 2011 to Sept 2011
Development of a large scale vector autoregressive time series estimator that runs in
the cloud. Produced with Hadoop, Cascading and Python. Built to handle hundreds
of thousands of time series that do not fit into memory. Implemented distributed
gradient descent that is capable of minimizing any differentiable objective function.
1 of 4
Belvedere Trading,
Chicago, IL
Software DeveloperJune 2011 to July 2011
Develop tools for algorithmic trading. During the academic year I developed RFi-
nance, an R package to do financial analysis, including Black Scholes option pricing,
the minimum variance portfolio, and a backtesting engine to measure the perfor-
mance of different trading strategies.Cisco Systems, San Jose, CA
Software DeveloperJune 2010 to Sept 2010
Cisco Choice intern in Physical Security. Designed and produced an automated
policy engine which intelligently scans a physical system for certain security condi-
tions, searches those conditions in a graph, and takes appropriate actions. Policy
Engine is highly optimized and parallelized. Wrote graph algorithms for connecting
a cluster of cameras and used the graph for shortest escape paths and optimizing
the deployment of a smart video search algorithm. Two patents pending.
Teaching Illahee Middle School, Federal Way, WA
Experience
Program DirectorSept 2006 to June 2007
Pioneered a new math class focusing on problem solving skills, mathematical intu-
ition, and creativity in math. Proposed course outline to school principal and school
district administrators.
InstructorSept 2006 to June 2007
Lead instructor for new math olympiad problem solving class. Wrote a new math
textbook and accompanying workbook focusing on math olympiad topics. Taught
two classes of 22 students everyday for the 2006-2007 school year.
Math Team CoachSept 2006 to June 2007
Created and coached a new afterschool math team of 100+ students. Team qualified
for MATHCOUNTS state competition in 2007.
Stanford University, Stanford, CA
Statistical Consulting Jan 2011 to Present
Consultant for the free Statistical Consulting service sponsored by the Department
of Statistics. Hold o ce hours every week where clients can drop-in and ask ques-
tions about experimental design, models, computational problems, data analysis,
and interpretation.
Education Program for Gifted Youth June 2008 to Aug 2008
Residential counselor and teaching assistant for number theory and math olympiad
problem solving. Lead a three hour section everyday for gifted high school students.
Awards Edmund F. Maxwell Scholar
Raytheon MathMovesU
Software R Packages
tsclust
GPU accelerated time series clustering using dynamic time warping. CUDA imple-
mentations of kmeans, kNN and hierarchical clustering.
2 of 4
fastVAR
Fast implementations for vector autoregressive estimation with exogenous inputs.
Diagnostics and prediction functions are available, as well as functions to use lasso
penalty for very large systems. Public Hadoop application to appear on Opani soon.
imputation
Missing data imputation using SVD, kNN, and SVT.
RFinance
Financial analysis including Black Scholes option pricing, minimum variance portfo-
lio, and a backtesting engine to test the performance of different trading strategies.
Plot the volatility smile and surface.
OptimalTrading
Extension for RFinance. Given a utility function, uses quadratic programming to
maximize returns while constraining risk. Given historic data, finds the optimal
trading strategy and allows users to benchmark their trading strategy. Uses the
apriori algorithm to find trading rules, e.g. buy X whenever you buy Y.
LargeRegression
Fast technique to solving large least squares problem using OLS and L2 normaliza-
tion.
MatrixBlockInverse
Fast technique to inverting a large matrix. Performs faster than the native R func-
tions. Functions provided for GPU acceleration for extremely fast performance.
Python Modules
mtsdb
Extracts time series data from a MySQL database, preprocesses it, and formats it
into a well constructed multivariate time series matrix. Used when analysis requires
studying multiple time series during the same time period and on the same time
scale.
Miscellaneous
PredatorPrey
Simulation for a pursuit evasion algorithm. Players are either predators, or prey,
which move on a 2D tile engine. Predators pursue prey, while prey attempt to get
to an exit and evade predators. Uses Java concurrency and graph algorithms.
BattleFieldChaos
Combat engine for an RPG game using Microsoft XNA.
Papers In Progress
GPU Accelerated Time Series Clustering using Dynamic Time Warping
RFinance: A Free Financial Analytics Tool
Projects In Progress
Data Driven Trading
Financial analytics platform. Developing a suite of software to help amateur in-
vestors become better informed about their investments. Want to bring tools only
available to hedge funds to small investors. Real time analysis of the markets, data
visualization, portfolio analytics, backtesting engine, time series models.
3 of 4
Rescue MAVs
Cooperative AI for micro air vehicles that will serve as support in search and rescue
missions. Programming intelligent team-based search algorithms.
Energy Analytics Dashboard
Developing a web app to help illustrate energy usage. Backend uses MySQL server,
frontend uses javascript and D3 for visualizations.
4 of 4