Sisi Zhang

M.S. Student, Khoury School of Computer Science, Northeastern University

Scalable Distributed System on AWS: Stores Ordering System

  • Design a RESTful, tomcat-based, scalable distributed web application that allows stores to make HTTP POST requests to record purchase orders and GET requests to look up previous orders.
  • Implemented in Java language, easily deployable to multiple AWS EC2 instances or run locally.
  • Built a multi-threaded client which simulates multiple stores, sends POST/GET requests to the servers and records QPS (query per second), mean/p95 latency of HTTP requests for performance evaluation.
  • Implement and test multiple strategies to scale-up the system, including multi-threading, AWS Load Balancer, Amazon MQ (with self-implemented connection pool), Amazon RDS, Amazon DynamoDB, Amazon DocumentDB (MongoDB). The system is deployed on 3 AWS EC2 instances, and passes a stress test of up to 1 million requests with about 6000 QPS.
  • Codes: github.com/zhangsisi0717/StoresOrderingSystem

Web Application: the Itinerary Planner

  • Visit at: zhangsisi0717.github.io/planner
  • The Itinerary Planner is a modern-looking, interactive web application built with Vue.js, powered by Google Maps APIs and Google Cloud Firestore.
  • Given origin / destination and a list of stops, it uses highly efficient algorithms for ATSP (asymmetric travelling salesman’s problem) and outputs a near-optimal itinerary that begins at the origin, visits all stops and ends at the destination using minimal travel time.
  • Features: interactive map with traffic map / satellite view, address auto-complete, real-time traffic-based travel time look up powered by Google Map APIs; sign up account, log in, save and retrieve itineraries, powered by Google Cloud Firestore.
  • Codes: github.com/zhangsisi0717/Optimal-Itinerary-Planner

Machine Learning and Data Mining in Mass Spectrometry

  • Propose a novel approach which applies non-negative Lasso regression to analyze and annotate large scale MS data.
  • Implement a efficient quadratic programming solver using Python to perform Lasso regression. Also builds a fully object oriented wrapper class for MS data that provides many useful helper methods in Python.
  • Achieves the state of art accuracy on public MS datasets.
  • Codes: github.com/zhangsisi0717/unintentional_fragments_annotation