Yin-Tung Chen

A highly passionate engineer from National Cheng Kung University, constantly seeks out innovative solutions to real-world problems and challenges.

  Tainan City, Taiwan          

Technical Skill

Software Engineering


3+ years experience in software development with modern frameworks.

  • Front-end: React Natvie, Vue.js, React.js, Cordova, Webpack, Babel, Expo, Google Analysis.
  • Back-end: Node.js (Express.js), PHP (Codeigniter), Python (Flask),  Golang (gin, gorm)
  • Database: MySQL, Postgres, SQLite, MongoDB, Firebase, Database Normalization.

Infra Engineering


2+ years experience in system design & server infrastructure engineering.

  • System Management: Linux user management (LDAP), resource control (cgroup).
  • Network Admin: DNS, NFS, TLS, web server (nginx), OSI.
  • Container: Docker (composing, networking, container registry), Kubernetes, VM management.
  • Cloud Platform: GCP (VMs, GKE), AWS (EC2, ELB, S3)

Machine Learning


1 years experience of machine learning theoretical & applicational techniques.

  • General: NN, RNN, LSTM, Attention mechanism.
  • NLP: Tokenize, Word embedding, S2s, Transformer, BERT.
  • Computer Vision: CNN, 
    VGG16, Resnet50.
  • Framework: Pytorch.

Work Experience

AInimal Team  •  Full-Stack Engineer & Product Manager

May 2020 - Present

AInimal is a start-up team consisting of National Cheng Kung University students, hoping to solve the community interaction troubles that young people encounter today, and create a healthy and friendly online dating environment.

Key tech: real-time message system, restful API, React Native, machine learning, cloud deployment.

  • Built this app into production and released on Android & iOS app store, currently having 9000+ registered users and 50+ daily active users.
  • Constructed a restful API  back-end service consisting of a three-layered architecture with Node.js & TypeScript stack and designed a 3rd normalized schema for its relational database.
  • Created an ORM-based abstracting layer for automatically migrating database schema in dev & prod environment with version control system, which free our developers from manually manipulating low-level database operations.
  • Constructed a real-time message service with Golang based on websocket protocol with a high-performance NoSQL database, which can serve up to 300 concurrent user connections.
  • Introduced DevOps pipelines using GitLab CI/CD for automatic unit testing & service deployment and thus successfully shortened development cycle time by 50%.
  • Designed a set of unit test suites on the API service layer based on Jest (a popular node.js unit testing framework), which improves the defect escape rate of production by 40%.

SideeX Team  •  Product Contributor

February 2020 - Present

SideeX is a professional automated record-and-playback web-testing tool with superior performance on its accuracy and efficiency, which is now maintained by members of the Software Engineering and Intelligent Test Automation Lab at National Cheng Kung University.

Key tech: software testing, selenium webdriver, web extension, electron framework.

  • Implemented the testing coverage feature on SideeX Runner, which significantly reduced manual efforts on creating test cases.
  • Conducted a research project consisting of the idea of testing coverage & web crawler and published as a conference paper, which was accepted by the IEEE ICEBE 2022 committee.
  • Merge this feature back to SideeX core, which is a large-scale project with up to 100K+ lines of code and 30+ cooperating contributors.

NCKU Computer & Network Center    System Developer

July 2019 - July 2021

Information System Development Division aims to develop the computerization for the respective requirements of different divisions at National Cheng Kung University, to improve administration efficiency and advance teaching services, including system planning, development, maintenance, management.

Key tech: MVC web service, PHP, C language programming.

  • Constructed web-based online system for official documents submitting, delivering and approving procedures for organizations at National Cheng Kung University.
  • Developed programs to simulate the university admission procedure and produce a correct result according to the admission rules established by the University Admission Committee in Taiwan.

Education

National Cheng Kung University  •  Step. 2018 - 2022

B.S. in Computer Science and Information Engineering

  • Avg. GPA: 3.99/4.30
  • Teaching assistant in Program Design (I) course 2019.
  • transcript:


Selected Project

AInimal App

The primary product of AInimal start-up team, with high-quality micro-service architecture and maintained by 10+ people software development team.

Used Techniques/Tools: React Native, TypeScript, Express.js, Socket.io, Redis, MySQL, FSM, PyTorch, MongoDB, Golang, nginx, Cloudflare CDN, Database migration.

Currently having 9000+ registered users, 50+ daily active users, and 60k+ messages.

     


CooMoney 

Final project of the User Interface Design & Development course

CooMoney is a web-based billing application. Different from common billing Apps, CooMoney is particularly designed for financial management of large teams with features such as separate billing projects, auditing mechanism, budget management.

Used Techniques/Tools: Vue, Vuex, Sass, Node.js, Express.js, MongoDB, mongoose.

Shortlisted in the final presentation of this course.

 
 


Transformer Chatbot

Independent Study

A deep learning model that can produce a response sentence with an input sentence given by users just like a real person chatting with you.

This model is based on "Transformer", a powerful sequence to sequence model that refers to the original paper "Attention is all you need". All components of this model are constructed using a pure Python program and Pytorch training framework without pretrained model API.



NearRest

Final project of the Theory of Computation course

Search for the best reviewed restaurants nearby you, add to your favorite list, view in Google Map. Line button carousel templates are used. Instead of typing a bunch of words into the LINE bot, you can just press buttons on the carousel template!.

Used Libraries/Tools: LINE messaging SDK, Flask, Heroku hosting, Firebase real-time database, Google map.





Mask Coverage Detector

Final project of Image Processing, Computer Vision and Deep Learning course

In this project, our model can automatically detect how many people are in a given photo, and the ratio of people who wear face masks properly among them. It can be used to indicate how dangerous this space would be.

We use FastRCNN predictor provided by torchvision as a pretrained model, and training data from kaggle face-mask-detection dataset.