I implemented this blog service featured with a modular and powerful editor. Welcome to signup and try it! If you think this is a good idea, you can also give donationopen_in_new to me through ecpay service.
Work on an interesting IoT service to help people to take care of their eye health. The challenges included
An experimental blog service shipped with pretty WYSIWYG editor and comprehensive social integrations.
Kickstarted a service of a startup company from scratch, including frontend development, backend development, version control, and operation.
Migrated, maintained and upgraded the backend of the website of 旺仔俱樂部 from Asp.Net to Nodejs.
2014.10 - 2015.07
Launched an open source project Seedopen_in_new, which helps developers speeding up the process of developing a website with Express and Angular.
2014.08 - 2016.07
Built the admin and management systems with python and django.
Developed the backend website of 新北市生技產業發展聯盟.
2013.09 - 2013.11
Developed the backend and the architecture of the website service and product.
2013.07 - 2014.08
Implemented social login system for the admin system of ezCloud service in PHP.
2016.09 - 2018.08
With the rapid advances of machine learning algorithms and sensing technologies, machine prognostics and health management (PHM) via data-driven approaches has become a trend in sophisticated machine tool industry. Recurrent neural network (RNN) is an important technique to process the sequential sensing data. Two popular PHM approaches are: 1) predicting the remaining useful life (RUL) of a machine, and 2) applying anomaly detection (AD) on a machine. Run-to-failure data is necessary to train RUL-based models. During the process of collecting run-to-failure data, a machine may face the situation of abnormality, such as impulse signals, before it fails. Therefore, AD-based approaches, which aim to detect anomalies in real-time, are more suitable for machine tool industry compared to RUL approaches. However, due to vanishing gradient problem in RNN, it is difficult and time-consuming to train an RNN model. To this end, we propose a framework, called QUAntized Recurrent neural network autoencoder for Time-series anomaly detection (QUART), to accelerate the RNN model training for the abnormality detection. The main idea of QUART is to reduce data size by mapping raw data into finite classes according to the probability mass in the raw data. To learn the anomaly states without considering the sequence order of sensing values in a short interval, we adopt a data shuffling method to increase the variance of the training data. Our experiment results show that the proposed framework can accelerate almost 120 times faster than the existing RNN methods.
2016.09 - 2018.08
Institute of Network Engineering, College of Computer Science, National Chiao Tung University, Taiwan, R. O. C.
2012.08 - 2016.06
Department of Computer Science (Network and Multimedia Engineering Program), National Chiao Tung University, Taiwan, R. O. C.
2014.06 - 2014.08
Taught HTML and CSS to 2 students.
2010.06 - 2011.06
Introduce from scratch basic python syntax, patterns and case studies to students.
Present my personal ongoing research on smart manufacturing.
Python, FastAPI, SQLAlchemy, Node.js, Express.js
React, Redux, Next.js, Mocha
PostgreSQL, MySQL, MongoDB
Circle CI, Travis CI
GCP, Docker, Kubernetes, Heroku
Deep Learning (mainly RNN, recurrent neural networks), Numpy, Pandas, Tensorflow, Keras
Git, Github, Gitlab, Jira, Clickup, Slack
Please refer to portfolio page.