KAI JU LIU 120 Ridge St., Apt. 3B New York, NY 10002 | (646) 541-4957 | kaiju@cims.nyu.edu PROFILE Excellent research and analytical skills. Dedicated to building creative and efficient software solutions for all applications. Strong programming track record in a wide range of standard languages (e.g. C/C++, Java) and scientific packages such as Matlab and Mathematica. Strong leadership and interpersonal skills; excellent verbal and written communication skills. EDUCATION New York University, Graduate School of Arts & Science, Ph.D., Mathematics, January 2006. GPA: 3.8 Stanford University, B.S., Mathematics, cum laude, June 2000. GPA: 3.7 DISSERTATION “A Trainable Graph Combination Scheme for Belief Propagation” Bayesian Statistics Studied Bayesian belief propagation, an efficient method to calculate probabilities in complex networks Designed and implemented novel scheme to apply belief propagation to networks with traditionally intractable loops Generated computer vision test case against optimal algorithm and alternative energy-approximation schemes Demonstrated new scheme competitive in accuracy and superior in complexity to other algorithms Object Recognition Applied new scheme successfully to distinguish geometric shapes from background in noisy images Devised learning algorithm to train thousands of scheme parameters for high-complexity problems Trained scheme on pre-processed images to successfully recognize animals in natural wildlife images TECHNICAL SKILLS General Programming Skills: C/C++, Java, Lush, HTML Scientific Programs and Languages: Matlab, Mathematica, LaTeX Computer Vision, C++, doctoral dissertation research (see DISSERTATION above) Developed multiple graphics IO and utility classes as well as specialized belief propagation classes (3,000+ lines) Managed large-scale computer vision tests and analyses spanning several days (2,000+ lines) Optimized algorithms for speed, memory efficiency, precision, and complexity Dynamic Animation and Control Environment (DANCE), C++, Visual Modeling final project Activated fully-articulated skeletal model with sixteen joints and thirty-seven degrees of freedom Created new modules to simulate balance, muscles, collisions, and angular friction Animated physically accurate skeletal pull-up under force of gravity Marble Madness, Java, Computer Graphics final project Re-created classic Marble Madness game as online interactive Java applet Built physics simulator and collision detector to model object interaction, gravity, and other forces Implemented ambient, diffuse, and specular lighting; phong shading, and kinematic motion PUBLICATIONS Voronoi Polyhedra of Unit Ball Packings with Small Surface Area. Károly Bezdek, Endre Daróczy Kiss, and Kai Ju Liu. Periodica Mathematica Hungarica 39.1-3 (1999): 107-118. WORK EXPERIENCE Instructor, September 2004 – December 2005, New York University, Calculus I and Calculus II Taught single-variable calculus in team of four: composed bi-weekly lectures and assignments; wrote and graded exams Researcher, Summer 1999, Cornell University, Research Experience for Undergraduates Studied and implemented new computational model for sphere packing (see PUBLICATIONS above) HONORS National Science Foundation Graduate Research Fellowship, September 2000 – May 2004 Stanford University President’s Scholar, September 1996 SKILLS AND INTERESTS Fluent in spoken Mandarin, proficient in German and written Mandarin Avid rock climber, runner, and tennis player