Education
- GPA: 3.88/4.30
- Core Courses: Computer Vision (A+), Machine Learning (A), Language and Statistics (A), Multimedia DB and Data Mining (A-), Search Engines (B+)
- GPA: 3.62/4.00
- Core Courses: Operating Systems (A),DataStructure (B+), Introduction to Database Systems (A), Compilers (A-), Software Engineering (A-), Introduction to Artificial Intelligence (A)
Work Experience
- Intern Project:
- Designed and developed Access-Pattern-Driven Disk-Level Block-Based Prefetch System as project leader
- Constructed adaptive prediction model with weighted correlation network
- Increased cache hit rate from 5% to 90%, and decreased remote system booting time from 8 min to 4 min
- Patent application is submitted by Riverbed Tehcnology
- Manager: Joshua Berry
- Intern Project:
- Designed and developed Data Distribution Network Visualization and Configuration Tool as project leader
- Parsed and formatted about 3,000 hand-written configuration files with flexible grammar
- Developed a GUI tool to generate the configuration files in uniform format
- Visualized data distribution network with D3JS, detected configuration errors including loops, undefined parameters and setting inconsistency
- Won Most Innovative Project among 57 Intern Projects
- Manager: Junjie(James) Qu
Research Experience
- Description: Use statistical, computer vision and deep learning methods to break 12306-Style CAPTCHA
- Supervisors: Tai Sing Lee, Christos Faloutsos
- Description: Supported SQL expression system and made some detective work on concurrent B-Tree, and wrote a part of automatic benchmarking scripts
- Supervisors: Anthony Tomasic, Andy Pavlo
- Description:
- Collected about 1,000 hand-written symbol forms, wrote a script to extract all individual symbols, and trained a mathematical symbol recognition model
- Developed a formula recognition algorithm by aggregating pixel blocks into mathematical symbols and parsing the formula structure with relative positions of symbols
- Developed an algorithm to correct recognition errors according to context
- Reached over 90% recognition accuracy in hand-written mathematical symbol recognition
- Supervisor: Deva Ramanan
- Description: Published Teaching Video Analytics Based on Student Spatial and Temporal Behavior Mining on ICMR 2015
- Supervisor: Hong Lu
- Description: Developed an algorithm to detect spam users and reduce their impact to rating scores
- Supervisor: Chaofeng Sha