Summary

Hi, I am Chun-Hsiang Wang, who is a software engineer at Google working on Google Cloud AI Platform now. Prior to this, I graduated from UC San Diego majoring in Computer Science in Dec 2019. Moreover, I joined Google as a software engineering intern in summer 2019 working on Kubeflow, an open-source machine learning toolkit for Kubernetes. I am excited about backend development, information retrieval, and natural language processing.

Welcome to contact me via emails.


Contact Information


Work Experiences

Google, Sunnyvale, USA

Software Engineer, Apr 2020 - Present

  • Working on Google Cloud AI Platform Online Prediction Service

Google, Sunnyvale, USA

Software Engineering Intern, Jun 2019 - Sep 2019

  • Improved the general-purpose label bot with 200+ users and 700+ repos by creating MLP classifiers for transfer learning to build personalized GitHub issue labelers and automated model training pipelines using Kubeflow
  • Designed the end-to-end bot to do repo-specific label prediction, handled GitHub requests asynchronously using Cloud Pub/Sub on Google Cloud Platform and deployed issue label services in production on Kubernetes
  • GitHub repository: kubeflow/code-intelligence & machine-learning-apps/Issue-Label-Bot
  • Constructed a new preprocessor to convert Jupyter notebooks to executable programs with Google Python-Fire (Kubeflow/Fairing)

National Chengchi University, Taiwan

Research Assistant @ CFDA & CLIP Labs, Aug 2017 - May 2018

  • Improved the accuracy ratio of the traditional financial model by 18% for the next 60 months by implementing LSTMs with the custom pairwise ranking loss to do time-series prediction using TensorFlow
  • Built a framework applicable to reviews in all domains to automatically generate domain-specific sentiment lexicons by applying representation learning and opinion mining using Python (Publication [1] in AAAI’19)
  • Paper: UGSD
  • GitHub repository: cnclabs/UGSD

Hewlett Packard Enterprise, Taiwan

Software Engineering Intern, Jul 2016 - June 2017

  • Speeded up server deployment 3X faster and initial setup 22% faster over previous generations by developing RESTful APIs and web applications
  • Increased code coverage by 30% by executing unit tests, set up testing environments using Docker and implemented 100% automated end-to-end tests plus test result reporting on Jenkins for CI/CD

Education

University of California San Diego, USA

Master of Computer Science, GPA: 3.79/4.00, Sep 2018 - Dec 2019

National Chengchi University, Taiwan

Bachelor of Computer Science, GPA: 3.98/4.00, Sep 2013 - Jun 2017


Projects

Recommender System, Apr 2019 - Jun 2019

  • Built translation-based recommender systems to do sequential prediction
  • Improved the current state-of-the-art models by 1.2% of AUC on average in 15 real-world datasets

CYK Parser, June 2019

  • Designed a binarizer doing structural annotation and vertical/horizontal Markovization for learning a probabilistic context-free grammar (PCFG)
  • Implemented CYK parsing algorithm to compute Viterbi trees and build a generative parser using Java, which achieves 80.2 of F1 score

Machine Translation, June 2019

  • Implemented two translation models, IBM model 1 and 2, and applied these models to predict English/Spanish word alignments
  • Designed an approach for growing alignments and improved the F1 score of IBM model 2 by 16.26%

Sequence Tagging, May 2019

  • Built a n-gram Hidden Markov Model (HMM) by implementing the Viterbi algorithm to tag gene names in biological text using Python
  • Improved the baseline tagger with the best one, 7-gram HMM, by 80.32% of F1 score

Language Modeling, Apr 2019

  • Implemented n-gram models with Laplace smoothing in Python and Kneser-Ney trigram language models in Java
  • Designed open address hash maps to do context encodings using Java
  • Generated sampled sentences from the learned probability distribution of words

Movie Recommender System, Nov 2018

  • Learned a Naive Bayesian model of movie ratings and adopted EM algorithm to build a personal movie recommender system

Speech Recognition, Nov 2018

  • Implemented the Viterbi algorithm to decode a discrete Hidden Markov Model (HMM) for automatic speech recognition (ASR)

Hangman, Oct 2018

  • Implemented a Hangman game that considers the belief network and automatically finds the best next guess according to the evidences by Bayes rule in Python

Brainstorming System, Jul 2016 – Jan 2017

  • Designed a retrieval-based intelligent system to facilitate real-time group brainstorming in Chinese online
  • Adopted text clustering and sentiment analysis by K-Means and Naive Bayesian with Scikit-learn, extracted keywords by TextRank, and produced word embeddings by Word2Vec using Python
  • Designed a web application to support chatrooms for more than 10 users using Node.js
  • Finished this work while I was an undergraduate research assistant at the Intelligent Media Lab, NCCU, funded by Ministry of Science and Technology (MOST), Taiwan (Publication [2] in ICMLC’18)

Motion Planner, Mar 2016 – Apr 2016

  • Designed autonomous polygonal robots to find collision-free paths in the 2D map containing obstacles from the starts to the goals and showed the robot actions on the GUI using Java
  • Built an artificial potential field with NF1 (local-minima-free) algorithm and C-obstacles for collision detection, and searched the C-space by implementing Best-first search

Chinese News Search Engine, Nov 2015 – Dec 2015

  • Created a search engine for more than 10000 Chinese news
  • Developed the backend to support the Boolean retrieval, the proximity search and the TF-IDF ranking

Travel Planning Site, Feb 2015 – Apr 2015

  • Designed a web application to show recommended travel itineraries by utilizing Google Maps API
  • Created pages using Parse and Facebook API to help users sign in, add their favorites and leave reviews

Open Source Contribution

Instagram Bot, [instabot]

  • Enhancement: allow Python API wrapper to download multiple photos in a post

Publications

  1. Chun-Hsiang Wang, Kang-Chun Fan, Chuan-Ju Wang, and Ming-Feng Tsai, “UGSD: User Generated Sentiment Dictionaries from Online Customer Reviews,” in Proceedings of the 33rd AAAI Conference on Artificial Intelligence (AAAI’19), Honolulu, 2019. [paper][code][slides]
  2. Chun-Hsiang Wang, and Tsai-Yen Li, “Design of an Intelligent Agent for Stimulating Brainstorming,” in Proceedings of the 10th International Conference on Machine Learning and Computing (ICMLC’18), Macau, 2018. [paper]

Technical Skills

Programming:

  • Python, C/C++, Shell Script, Java

Frameworks and Tools:

  • Kubernetes, Docker, Google Cloud Platform, Kubeflow, Apache Airflow, Apache Kafka, TensorFlow, Jenkins, Linux/Unix, Git