17015048_1737902523188954_27829206_o.jpg

Chiew Jun Chuan

Data Scientist | Artificial Intelligence Engineer | Deep Learning Enthusiast


Working Experience

Computer Vision Engineer | Algorithm Engineer

Viscovery , 2018/05/02 - Present

A company that providing comprehensive recognition solutions for Smart Retail, Smart E-Commerce, Smart Media, Augmented Reality, Smart Factory and more.

  • Smart-checkout solution: a vision base merchandise / bread checkout system that applied to convenience stores, bakery stores, POS vendor, etc.
    • object detection & instance segmentation for located foreground objects accurately.
    • coarse / fine grained classification for recognize objects accurately.
    • metric / representation learning for improve efficiency of new product / unknown product recognition.
    • data augmentation & generation for reduce cost of data collection.

Data Scientist | Artificial Intelligence Engineer | Research Assistant

Academia Sinica | Institute of Information Science (IIS) | Data Insights Research Lab (DIRL) | Taiwan AI Academy , 2016/06/27 - 2018/04/30

A lab provided machine learning & deep learning solutions for our business partners.

  • Text mining and the prediction model for book sales (with books.com.tw)
  • Exploratory data analysis and the linear model for the KPI of marketing events (with EASYCARD Co., Ltd.)
    • handle more than 100 millions data.
  • Exploratory data analysis for the passenger boarding behavior on MRT (with EASYCARD Co., Ltd.)
    • handle more than 100 millions data.
  • Time series classification using deep learning for the prediction market on soccer
  • Data mining for the players gambling behavior
    • handle more than 6 millions bets and 3,000 matchs soccer events, approximately 100GB of data.
  • Exploratory data analysis and the prediction model for the customer satisfaction (with HELLO TAIPEI)
  • Text mining and the prediction model for SMS copy design (with Taiwan Mobile Co., Ltd.)
  • Churn rate prediction model (with Taiwan Mobile Co., Ltd.)
    • handle more than 6 millions users.
    • get a 0.81 AUC performance in prediction model.
  • Dye selection and concentration prediction model with deep learning (with Everest Textile Co., Ltd.)
    • get a 92% top-3 accuracy & 99% top-10 accuracy in selection model.
    • get a 0.99 correlation in concentration model.
  • Project Theta, the project for the Taiwan industry upgrade with artificial intelligence (with Advantech, Chang Chun Group, Formasa Petrochemical Corporation, etc.)
    • Task included automatic optical inspection (AOI) , parameter optimization, predictive maintenance, etc.

Internship

Fortuna Intelligence Co., Ltd. , 2016/03/01 - 2016/06/24

The best digital brokerage firm for self-directed investors. Currently launched the main service: Fugle, a quick and concise securities research space; and facebook messenger bot.

  • Automatically record the date of company's investor meetings with web crawler and google calendar API
  • Automatically simplified financial information major message
  • Stock market data analysis & user behavior analysis

Education

Bachelor of Science, Risk Management and Insurance

National Chengchi University, 2012/09/01 - 2016/06/30

  • Major: Actuarial Science and Statistic

Skills


Programming

  • Python (familiar in numpy, pandas, flask, etc.)
  • Computer Vision (opencv, pycocotools, imgaug, etc.)
  • Machine learning (familiar in keras, tensorflow, etc.)
  • R (familiar in data.table, ggplot2, etc.)
  • Working Environment (vim, tmux, docker, git)


Text Mining

  • Topic model 
  • Word Embedding (familiar in gensim, text2vec, etc.)


Data Analysis & Machine Learning

  • Exploratory data analysis & data mining 
  • Statistical analysis & simulation 
  • Feature engineering 
  • Prediction modeling (familiar in XGBoost, LASSO, etc.)


Deep Learning

  • Metric learning (image embedding) / representation learning
  • Component selection optimization (multilabel classification)
  • Time series, text classification with convolutional neural network and recurrent neural network
  • Predictive maintenance & anomaly detection with deep neural network
  • Image (defect) classification with convolutional neural network