Quan Li |

Ph.D Candidate, Electrical Engineering and Computer Science, Massachusetts Institute of Technology, MA, USA, Sep 2012 - present

GPA: 5.0/5.0

Minor in Quantitative Finance. GPA: 5.0/5.0

Thesis: Algorithms and Algorithmic Obstacles for Probabilistic Combinatorial Structures

Committee: Prof. Guy Bresler, Prof. David Gamarnik (advisor), Prof. Caroline Uhler.

M.S., Applie Math in Control and Dynamical Systems, Duke University, NC, USA, Sep 2010 - May 2012

GPA: 4.0/4.0

M.S., Civil Engineering, Tsinghua University, Beijing, P.R. China, Sep 2007 - Jun 2010

GPA: 91.3/100

B.S., Civil Engineering, Tongji University, Shanghai, P.R. China, Sep 2003 - Jun 2007

GPA: 87.4/100

Design and Analysis of Algorithms

Machine Learning

Statistics

Data Mining/Analysis

Probability

High-performance C/C++ code and python vectorized code

KCG Holdings, Inc/Virtu Financial.

Quant Intern

As a quant intern in its Client Market Making Group, I improved the speed and functionality of its internal client market making trading simulator, did research on the price impact module of the simulator, as well as partially implemented this module. The code was written in Python/Cython/C++.

Celect, Inc. Jun 2016–Aug 2016 Boston, MA

Data Scientist/Algorithm Developer Intern in a start-up based out of MIT.

Supervised by Prof. Devavrat Shah

Design, implement and test various collaborative filtering algorithms (matrix completion, tensor completion, Feature-based inference, etc.) for Celect recommender system. The code was written in Python/Cython/C++.

Presidential fellowship-Joan and Irwin Jacobs Fellowship, MIT, 2012

Multiple travel grants, 2014-2015

Workshop for Graphical Models, Statistical Inference and Algorithms, Institute for Mathematics and its Applications (IMA), University of Minnesota.

Summer school for Graph limits, Groups and Stochastic Processes, The Alfred Renyi Institute of Mathematics, Budapest, Hungary.

Best Master Thesis Award (Select for the best master thesis among 101 graduate students), Tsinghua University, 2010

China State Construction Engineering Research Award, Tsinghua University, 2010

Highest-level Scholarship for Comprehensive Excellence, Tsinghua University, 2008-2009

National Undergraduate Contest of Mathematical modeling, Second Prize (team leader), Tongji University, 2005-2006

*“Finding a Large Submatrix of a Gaussian Random Matrix”, To appear in Annals of Statistics (accepted), arxiv:1602.08529 (with D. Gamarnik)

*“Matrix Completion from O(n) Samples in Linear Time”, The 30th Annual Conference on Learning Theory (COLT 2017) and To be submitted to Annals of Statistics (long version), arXiv:1702.02267 (with D. Gamarnik and H. Zhang)

*“On the Max-Cut of Sparse Random Graphs”, To appear in Random Structures & Algorithms (accepted), arxiv:1411.1698 (with D. Gamarnik)

Q Li, JT Scruggs. Asymptotic Analysis of Vector ARMA Identification, IEEE Conference on Decision and Control (CDC ’12), Maui, Hawaii, USA, (2012).

JT Scruggs, Q Li. Passive Network Design for Stochastic Vibratory Energy Harvesters, IEEE Conference on Decision and Control (CDC ’11), Orlando, Florida, USA, (2011).

Q Li, JS Fan, JG Nie, QW Li. Crowd-induced Random Vibration of Footbridge and Vibration Control Using Multiple Tuned Mass Dampers, Journal of Sound and Vibration, 329(19), 4068-4092 (2010).

JS Fan, JG Nie, Q Li. Long-Term Behavior of Composite Beams under Positive and Negative Bending (II) - Analytical Study, Journal of Structural Engineering, ASCE, 136(7), 858-865 (2010).

Experiment on the financial market data for forcasting stock index direction/stock prices using various machine learning techniques, including Ridge Regression, LASSO, Elastic Net, Forward/Backward-stepwise Selection, Least Angle Regression, Principal Components Regression, K-Nearest Neighbors, Logistic Regression, LDA, QDA, Bagging, Random Forests, Gradient Boosting, Recurrent Neural Network (LSTM Networks), etc. The code was written in R and Python.

Final Project for Performance Engineering of Software Systems (6.172): Leiserchess. The project starts with a slow serial game-playing AI for Leiserchess-—a chess-like game with lasers. We speed up this AI by redesigning board / pieces representation using bit hacks, implementing parallel Principal Variation Search (NegaScout) which includes various optimization on synchronization with / without Locks, coarsening, caching, etc., implementing opening / endgame book, etc. Finally our bot achieves roughly 23 million nps and search depth 8 on average using 8 cores in Microsoft Azure. Finally tournaments were run among 26 bots contributed by 26 teams. Our bot was ranked 6/26 in the first submission and 4/26 in the second submission. The code was written in C/C++.

Clustering U.S. Census Data. We found demographic clusters in census data by building a categorical mixture model and fitting it using K-means, K-means++ and Expectation Maximization (EM) algorithms. Also, we trained the Gaussian Mixture Model (using EM) of the demographics for each state in U.S. Using these models, we developed a novel way of measuring the similarity of two states’ populations. The code was written in Python (Numpy and Pandas).

Serial Dynamic Memory Allocation. We implemented a binned free list structure to provide a memory management unit, which consists of four major functions: init(), malloc(), free(), realloc(). Due to a lot of untyped pointer manipulation which makes dynamic memory allocators notoriously tricky to program correctly and efficiently, we also write a heap checker that scans the heap and checks it for consistency. Finally, our memory allocator was evaluated by space utilization and throughput against a number of memory allocation traces. Although our implementation consumes slightly more space than the implementation based on the single free list, our implementation is much faster. The code was written in C/C++.

Collision Detection: optimize a graphical screensaver program for multicore processors using the Cilk Plus parallel programming interface. The code was written in C/C++.

Bit Hacks: improve performance of programs using the perf tool and experiment with word-level parallelism. The code was written in C/C++.

MIT Dept. of EECS. Sep 2016 — Dec 2016

Teaching Assistant for Machine Learning (6.867)

Lecturer: Tomas Lozano-Perez, Suvrit Sra, Leslie Kaelbling

Graduate core class in Artificial Intelligence for MIT CS PhD

MIT Dept. of EECS. Jan 2017–Present

Teaching Assistant for Introduction to Machine Learning (6.036)

Lecturer: Tommi Jaakkola, Regina Barzilay, Stefanie Jegelka, Pablo Parrilo

Undergraduate core class in Artificial Intelligence for MIT CS major

MIT Dept. of EECS. Sep 2013 — Dec 2013

Teaching Assistant for Advanced Stochastic Processes (6.265)

Lecturer: David Gamarnik

Advanced graduate class for MIT EECS PhD

C/C++, Python (Numpy and Pandas), Cython

Reviewer for Annals of Probability

Reviewer for Random Structures and Algorithms

Computer Science

Performance Engineering of Software Systems; Elements of Software Construction; Machine learning; Introduction to Deep Learning; Advanced Algorithms; Design and Analysis of Algorithms; Introduction to Algorithms; Cryptography and Cryptanalysis.

Math/Statistics

Theory of Probability (with heavy analysis); Advanced Stochastic Processes; Theoretical Statistics; Statistical Learning Theory and Applications; Mathematics of Machine Learning; High Dimensional Statistics; Functional Analysis; Real Analysis; Abstract Algebra; Asymptotic Analysis and Perturbation Methods; Mathematical Modeling; Scientific Computing; Mathematical Programming (Optimization).

Finance

Algorithmic Trading and Quantitative Investment Strategies (currently taking); Options and Futures Markets; Finance Theory; Analytics of Finance; Data Technologies for Quantitative Finance; Quantitative Investment Management.

Department of Electrical Engineering and Computer Science, MIT

77 Mass. Ave., 32-D572

Cambridge, MA 02139

Email: quanli at mit dot edu