Hi there! I am an incoming Ph.D. student in Computer Science at UC San Diego. I recently graduated with a B.S. in Computer Science and in Applied Math from UCSD (with Honors), luckily advised by Prof. Zhuowen Tu. My research interest lies in theory & applications of machine learning, with a focus on computer vision. This website contains my publications, selected work experience, and some small projects.
Software Engineering Intern
@ Google (June 2017 to Sept. 2017)
■ Designed a deep learning model that predicts the completion time of data rating tasks, used by dozens of teams at Google.
■ Reduced the average prediction error by 90% even with limited amount of training data (as data collecting is ongoing).
■ Implemented a predictor for the model and deployed it as a service that handles on-demand prediction requests.
■ Automatized the data collection and feature extraction process from different data sources (Spanner & Bigtable).
■ Performed data cleansing and feature analysis.
■ Integrated this functionality into front-end UIs.
@ Machine Learning, Perception, and Cognition Lab (Jan. 2016 to present)
■ Performing computer vision research.
■ Worked on an image labeling pipeline for a quad-copter 3D simulator.
@ Cardiac Mechanics Research Group (Sept. 2016 to May 2017)
■ Optimized computation codes written in a combination of Python, C++, CUDA and Fortran for Continuity, a large-scale electrophysiology simulation software.
Tutor and Grader
@ UCSD CSE department (Oct. 2015 to June 2017)
■ Tutored data structure (CSE 12) and machine learning (CSE 150).
■ Privately tutored Python programming and applied statistics.
An Android app as a mixture of calendar and todo-list, supportive of dynamic allocation of tasks with deadline. I designed and implemented a login system and a notification system with a web service in . View the project here.
A GPU-supported Neural Network library that consists of feed-forward and recurrent net with different implementation versions. Specifically built fully-connected neural nets with specially designed pooling layers and a long short-term memory. Achieved 99.02% testing accuracy on MNIST dataset, which is better than results of any fully-connected neural net. View codes HERE.
An online unit calculator. Built with a parser and an interpreter. Specifically, users can define new units, simplify expressions, and do calculations on numbers with units. Deployed with PostgreSQL for user-defined units. Try it here. Or view backend codes.
A classic combinatorial game. Easy rules and intriguing. I implemented the wining strategy algorithm for this game (without implementing a GUI, though) View source codes here.
A package for linear and nonlinear optimization, with a tool for solving nonlinear equations. I implemented Simplex Method for linear programming and modified Newton Method for nonlinear problems. Source codes are here and MATLAB / Matlab Compiler Runtime Evironment (MCR) or Octave is needed to run it.
A hardware emulator that takes assembly codes and runs the input programs. Designed for a specific kind of general-purpose assembly codes. Source codes are here.