chenxi1103
Software Dev Engineer @ AWS / Android, iOS development / Full-Stack Development / Big Data / Machine Learning
Amazon Web Services, IncSeattle
Pinned Repositories
Auto-Word-Completion
Realized Auto-Word-Completion by MapReduce/MySQL, visualized by PHP/Ajax/JQuery. Algorithm used: N-Gram Model
Carcassonne-Game-Software
Design and Implemented the Core and GUI of the Carcassonne Game with 5500 lines of Java Code individually. Realized and Demonstrated the Design by UML diagrams including Domain Model, Interactive Diagrams, System Sequence Diagrams and Object Model. Applied OOD design pattern including Strategy Pattern, and Observer Pattern. Fully test the Core part by JUnit test. Utilized Git, Gradle, and TravisCI to manage the project.
ChatRoom
ChatRoom realized by Django web socket, supporting inviting friend to chatroom by sending link through email.
chenxi1103.github.io
my personal technical blog
Coding-Practice
For saving the solutions to the coding questions
Face_Recognition_Project
Gender/Race/Emotion classifications based on facial multi-attribute detection were realized through data pre-processing, face detection and extraction by OpenCV, developing training models through “Fisherfaces” and “Convolutional Neural Network” approaches by TensorFlow and Keras with accuracy of 98.44%, 84.24%, and 70% respectively. A web app was developed for better demonstration and further model optimization by expanding training dataset and user labeling, adopted Nginx to serve web server and deployed on DigitalOcean.
Google-BigTable-Implementation
Implement a distributed Bigtable-like database which supports effective CRUD for large-scale data with one master server and multiple tablet servers. The master server directly communicates with clients through RESTful API and manage the requests by allocating the tasks among tablet servers like table creation and deletion on assigned tablet servers, locking the tables, sharding (splitting of a single table amongst multiple tablet servers for storage effectiveness), health-checking for tablet servers through heartbeat, recovery (handle the permanent loss of a tablet server) based on WAL (write-ahead log) and metadata, and start/kill the servers. The tablet servers support garbage collection (only keep the last five inserted values), maintaining the memtable for effective searching (memtable would spilled to disk when there are a hundred unique row keys exists), maintaining the in-memory indexes, sharding when a table contains more than 1000 row keys, WAL, reconstructing the tablet server state using WAL and SSTables maintained in the persistent storage, effective range searching in SSTables.
Netflix-Movie-Recommendation-System
Developed a full-stack movie recommendation system with RESTful API to provide clients with 20 recommended movies. Determine the hit rate by collecting clients’ watching data to see if they watch the recommended movies afterwards. Continuous Integration for pipeline code. Automated daily model quality evaluation and system supervision with Jinkens. Designed and built the infrastructure that can incrementally deploy new versions of recommendation service triggered by canary release and A/B testing. Integrated with feedback loops mechanism to detect potential positive or negative feedback loops to further identify potential adversarial attacks. Implemented the monitoring and detection by applying lambda architecture to combine the stream and batch processing results to detect problematic behaviors. Comprehensive data quality control on raw data received from Kafka stream, especially focus on data schema issues, missing data, and duplicated data. Monitoring Dashboard UI with D3.js. Developed the whole web server by Flask. Containerized the whole service by Docker.
SmartZillow
A real estate search and value prediction system using Service-oriented Architecture, visualized by web development through Node.js, Express, and Bing Map API. Developed the distributed web scraping system to collect property information through Python, MongoDB, and CloudAMQP (RabbitMQ). Designed and developed the estate value prediction system with TensorFlow.
Tamagochi-A-web-based-pet-simulation-game-
Best Final Project in CMU 17637. A web-based pet simulation game which supports register/change password through email, generate pets randomly among 32 characters, pet evolution (4 levels) based on login time (age), shopping, feeding, health care, adding friends, propose and marriage, love wall, death record, with single-player games which based on “flappy bird” and “Whac-A-Mole” were developed by JavaScript, and one multi-player game which allows fluid multi-user real-time interactions was developed by web socket, mainly developed by JavaScript/jQuery/Ajax/Django, adopted Nginx to serve web server, configurated and deployed on AWS EC2, implemented MySQL as database. Implemented with weather API to fetch and update the real-time weather every 5 seconds in Pittsburgh which influences the pet’s health, and geological API to fetch the locations of all users to display the “nearby” users (radius less than 5km) for adding friends.
chenxi1103's Repositories
chenxi1103/Netflix-Movie-Recommendation-System
Developed a full-stack movie recommendation system with RESTful API to provide clients with 20 recommended movies. Determine the hit rate by collecting clients’ watching data to see if they watch the recommended movies afterwards. Continuous Integration for pipeline code. Automated daily model quality evaluation and system supervision with Jinkens. Designed and built the infrastructure that can incrementally deploy new versions of recommendation service triggered by canary release and A/B testing. Integrated with feedback loops mechanism to detect potential positive or negative feedback loops to further identify potential adversarial attacks. Implemented the monitoring and detection by applying lambda architecture to combine the stream and batch processing results to detect problematic behaviors. Comprehensive data quality control on raw data received from Kafka stream, especially focus on data schema issues, missing data, and duplicated data. Monitoring Dashboard UI with D3.js. Developed the whole web server by Flask. Containerized the whole service by Docker.
chenxi1103/Face_Recognition_Project
Gender/Race/Emotion classifications based on facial multi-attribute detection were realized through data pre-processing, face detection and extraction by OpenCV, developing training models through “Fisherfaces” and “Convolutional Neural Network” approaches by TensorFlow and Keras with accuracy of 98.44%, 84.24%, and 70% respectively. A web app was developed for better demonstration and further model optimization by expanding training dataset and user labeling, adopted Nginx to serve web server and deployed on DigitalOcean.
chenxi1103/Coding-Practice
For saving the solutions to the coding questions
chenxi1103/Google-BigTable-Implementation
Implement a distributed Bigtable-like database which supports effective CRUD for large-scale data with one master server and multiple tablet servers. The master server directly communicates with clients through RESTful API and manage the requests by allocating the tasks among tablet servers like table creation and deletion on assigned tablet servers, locking the tables, sharding (splitting of a single table amongst multiple tablet servers for storage effectiveness), health-checking for tablet servers through heartbeat, recovery (handle the permanent loss of a tablet server) based on WAL (write-ahead log) and metadata, and start/kill the servers. The tablet servers support garbage collection (only keep the last five inserted values), maintaining the memtable for effective searching (memtable would spilled to disk when there are a hundred unique row keys exists), maintaining the in-memory indexes, sharding when a table contains more than 1000 row keys, WAL, reconstructing the tablet server state using WAL and SSTables maintained in the persistent storage, effective range searching in SSTables.
chenxi1103/Carcassonne-Game-Software
Design and Implemented the Core and GUI of the Carcassonne Game with 5500 lines of Java Code individually. Realized and Demonstrated the Design by UML diagrams including Domain Model, Interactive Diagrams, System Sequence Diagrams and Object Model. Applied OOD design pattern including Strategy Pattern, and Observer Pattern. Fully test the Core part by JUnit test. Utilized Git, Gradle, and TravisCI to manage the project.
chenxi1103/ChatRoom
ChatRoom realized by Django web socket, supporting inviting friend to chatroom by sending link through email.
chenxi1103/chenxi1103.github.io
my personal technical blog
chenxi1103/SmartZillow
A real estate search and value prediction system using Service-oriented Architecture, visualized by web development through Node.js, Express, and Bing Map API. Developed the distributed web scraping system to collect property information through Python, MongoDB, and CloudAMQP (RabbitMQ). Designed and developed the estate value prediction system with TensorFlow.
chenxi1103/Tamagochi-A-web-based-pet-simulation-game-
Best Final Project in CMU 17637. A web-based pet simulation game which supports register/change password through email, generate pets randomly among 32 characters, pet evolution (4 levels) based on login time (age), shopping, feeding, health care, adding friends, propose and marriage, love wall, death record, with single-player games which based on “flappy bird” and “Whac-A-Mole” were developed by JavaScript, and one multi-player game which allows fluid multi-user real-time interactions was developed by web socket, mainly developed by JavaScript/jQuery/Ajax/Django, adopted Nginx to serve web server, configurated and deployed on AWS EC2, implemented MySQL as database. Implemented with weather API to fetch and update the real-time weather every 5 seconds in Pittsburgh which influences the pet’s health, and geological API to fetch the locations of all users to display the “nearby” users (radius less than 5km) for adding friends.
chenxi1103/17645GroupProjectTeamA
chenxi1103/Auto-Word-Completion
Realized Auto-Word-Completion by MapReduce/MySQL, visualized by PHP/Ajax/JQuery. Algorithm used: N-Gram Model
chenxi1103/18755Project
Troll Farm CMU: Identifying and Manipulating the Aging of Social Networks
chenxi1103/cdn
jsDeliver CDN
chenxi1103/Content-Distribution-Service-P2P-Network
Designed and implemented a P2P discovery network based on link-state advertisement protocol, which supports dynamically peer discovery and generation of global network topology with priority path selection, via UDP-variant communication protocol.
chenxi1103/CPP_Learning
chenxi1103/discord-music-bot
chenxi1103/Google-PageRank
Realized PageRank by processing data by MapReduce, visualized by JS/JQuery/D3.js
chenxi1103/LeetCodePractice
Have no idea how long will it last but...
chenxi1103/MagicBBS
Standard Bulletin Board System Construction By J2EE
chenxi1103/Nano-Blogsite
Nano-Blogsite realized by Django2.1/ Python3.7/ HTML/CSS/JavaScript/JQuery/Ajax
chenxi1103/Objective-C-Learning
Objective-C learning from the beginning. Detailed notes included (in Simplified Chinese).
chenxi1103/P2P-Video-On-Demand-HTTP-Server
chenxi1103/seai
CMU 17-445/645: Software Engineering for AI-Enabled Systems