AUDREY LEONG


Hi, I’m Audrey, a Master’s student at Columbia in Computer Science, graduating Dec. 2025. My interests are at the intersection of ML, generative AI, and human computer interaction. I believe that new technologies  are here to stay, and want to explore how they can extend human capabilities.

Here are some of my projects, I’m happy to discuss any of them! 

Resume |  LinkedIn

ML/GenAI
1: NotePilot
2: YOLOv8 Traffic

Systems
3: Image Quilting

Game Dev
4: Hammered

Data Viz and Mining
5: Squirrels in the City
6: The Netflix Nexus

audrey.leong@columbia.edu
650.518.9528
New York, NY 

1: Notepilot

Supervised by Prof. Lydia Chilton
Fall 2024

This is a project exploring the inclusion of generative AI into design processes and everyday tools. NotePilot is build for university students, taking syllabi and homework assignments to build customized Notion templates. 

The focus is on creating individualized schedules for students to know the last possible day they can start an assignment and still make it to office hours for help.  Full Flask app integrated with OpenAI API.

Demo Video linked here



2: YOLOv8 for Traffic Light Recognition
Columbia Deep Learning for CV

Motivated by the widespread interest in autonomous vehicles and making them safer, this project benchmarks different traffic light datasets to determine YOLOv8 accuracy and speed in classification. Various data augmentations (brightness, rotation, mosaic tiling) were performed for model robustness.

Project report linked here.
Fall 2024 


3: Image Quilting Optimization
ETH Advanced Systems Lab

Systems project focused on improving a single-threaded image quilting algorithm in C via algorithmic optimizations, AVX-intrinsics, auto-tuning, and code generation. Final speedup achieved is 15x over the initial baseline. Optimized for an Intel Core i7 machine GCC 12.2 compiler. 

Final report linked here.
Summer 2023 


4:Hammered
ETH Game Programming Lab

An engaging 3D single-player game, Hammered is a puzzle game based on the concept of Thor’s magical hammer and its ability to always freturn to its owner. 

Developed entirely from scratch in Monogame, the game includes custom-made assets and music, a unique offline path-finding algorithm for 3D space, and dynamic puzzles for the player to enjoy!

Final report linked here and open source repository linked here

Final build is free to play here! (Windows only)
Spring 2023


5: Squirrels in the City
Columbia GSAPP

Done in Grasshopper and Rhino, this 3D simulation model of the New York City squirrels’ movements can illustrate their behavioral patterns throughout different times of day and seasons.

Based on the issues of coexistence between human and urban wildlife — that the squirrels’ foraging pattern largely overlaps with paths of human activity — this project aims to tweak human behavior by relocating human life facilities, such as trash cans, in order to better delimitate the squirrels’ area of activity and decrease the potential ‘aggressive’ interactions.

Full project report linked here.
Spring 2022


6: The Netflix Nexus
Columbia Data Visualization

Given Netflix’s Top 10 data, the first publicly available data since the deprecation of its API in 2014, the goal is to explore media coproducers and see if certain combinations contribute to the virality of Netflix media. Done in Observable d3.js. 

Interactive project linked here.
Project article can be read here.
Spring 2022