Software Engineer

Ge Shi https://geshijoker.github.io/

Mobile : 413-***-**** **********@*****.***

Education

University of Massachusetts{Amherst Amherst, MA

Master of Science in Computer Science; GPA: 3.90/4.00 May. 2019 Zhejiang University Hangzhou, China

Bachelor of Engineering in Automation; GPA: 3.65/4.00 Sept. 2013 { July. 2017 Programming Skills

Languages: Python, C/C++, Java, Matlab, SQL, Javascript, R, D3. Courses: Advanced Algorithm, Data Base, Distributed System, Statistics and Probability, Uncertainty Inference, Networks.

Abilities: Machine Learning, Computer Vision, TensorFlow, Database, Image Processing, Robotics(ROS), Deep learning, Data Visualization, Programming Principles.

Internship Experience

NetEase Huyu Incorporation, Limited Hangzhou, China Game AI Software Engineer Nov. 2016 - Feb. 2017

Classi cation: Engineered supreme supervised classi cation theory using Python, SQL and database technology; Classi ed the preferences of the game players for roles’ careers, equipments, ornaments and game modes to optimize recommendation system.

Behavior Trees: Took part in the design of game AI behavior trees of a ’guardian’, which based on the habits of player to patrol and track enemies; xed bugs to make AI robot behave normally.

Internet worm: Created and Exerted internet worms on the online interest sharing communities like Baidu Post Bar, Weibo etc. to get the information of interested potential customers, which facilitated the propagation and market department.

Huiying Medical Technology (Beijing) Incorporation, Limited Beijing, China Image Processing Software Engineer Jul. 2016 - Aug. 2016

Machine learning: Applied machine learning theory onto the auto-diagnosis systems of medical images to decide the focal position of DICOM image collected from the server via WADO, and outlined target area with the image cloud PACS system.

Image Processing: Developed the machine learning framework in C++ as a member of algorithm team which pertains to system development and mining algorithms for image segmentation and image registration. Academic Experience

Moving Object Segmentation Amherst, MA

Supervisor: Prof. Learned-Miller March. 2018 - May. 2018

Deep Neural Network: Adopted encoder-decoder model to preserve spatial information of pooled deep features. Based on the initialization of optical

ow net to transfer learning segmentation boundaries.

Unsupervised Learning: Leveraged well designed loss function to train Neural Networks with unsupervised learning. Based on recent papers, improved the architecture to train a more general model to well t optical

ow and reduce over tting.

Model Inference Processes Amherst, MA

Supervisor: Prof. Brun Oct. 2017 - Jan. 2018

Model Inference Techniques: Collected log les with multiple traces of human-driven process descriptions for the same process, then managed them into Node.js network type using JavaScript and customized model inference techniques (Synoptic) to infer an FSM ( nite state machine) model that captures all the variations in the process.

High Level Goal: Evaluated the generative properties of the model with JAVA code snippets. Performed sensitivity analysis on the inclusion of more or fewer traces. The high level goal is to build self-monitor system which is conducive to the stability of self-adaptive systems in dynamic environments. The result is visualized with Data-Driven Documents (D3) based on JavaScript online.

Contact this candidate