Oscar Shih

Data Science graduate student at the University of Michigan, skilled in machine learning, analytics, database systems, and quantitative modeling. Focused on building robust, scalable solutions that bridge theory with practice.

🗄️ Facebook-Style Relational Database Design

🔗 GitHub Repo

Description
This project involves designing and implementing a relational database schema for a social media platform inspired by Facebook. It encompasses the creation of tables, enforcement of data integrity through constraints, and development of SQL scripts for schema creation and data management.

  • Designed an ER diagram to model entities such as users, friendships, messages, photos, and events
  • Implemented SQL scripts to create and manage database objects, ensuring referential integrity
  • Developed views and queries to facilitate data retrieval and analysis
  • Included scripts for loading sample data and dropping database objects for maintenance

🖧 Facebook-Style JDBC Database Interaction

🔗 GitHub Repo

Description
This project demonstrates how to integrate Java applications with relational databases using JDBC. It focuses on executing SQL queries to retrieve and process data related to users, friendships, and events in a social network context.

  • Implemented SQL queries to extract meaningful insights from the database
  • Utilized JDBC for database connectivity and data manipulation
  • Ensured proper resource management by closing connections and statements appropriately
  • Organized code to separate concerns between data access and business logic

🗃️ MongoDB User Data Analysis & Query

🔗 GitHub Repo

Description
This project demonstrates how to migrate relational data to a NoSQL system and perform analytical queries using MongoDB.
It showcases data extraction from Oracle, transformation into JSON, and aggregation queries to reveal user and friendship patterns.

  • Extracted user data from Oracle and transformed it into MongoDB-compatible JSON documents
  • Imported data into MongoDB and structured collections based on users, friendships, and cities
  • Performed queries to analyze friend counts, restore broken friendships, and extract birth month distributions
  • Applied MongoDB’s aggregation framework for efficient data grouping and statistical analysis
  • Demonstrates practical skills in data migration, NoSQL modeling, and query design

📀 Grace Hash Join Implementation

🔗 GitHub Repo

Description
This project implements the Grace Hash Join algorithm to perform efficient joins on large datasets under memory constraints.
It simulates disk I/O behavior, applies multi-pass partitioning, and builds memory-efficient hash tables to probe and merge matching records.

  • Designed partition phase using a hash function to divide large relations into disk-based buckets
  • Developed probing phase with in-memory hash tables to join partitioned data efficiently
  • Handled recursive repartitioning when intermediate partitions exceed available buffers
  • Simulated buffer pool management and disk I/O to mimic realistic system limitations
  • Demonstrates practical understanding of external join algorithms, buffer optimization, and query execution planning