Marco Yang

Experience

Machine Learning Researcher | Caltech Vision Lab

Mar 2024 - Dec 2024

Novel approach to cross-domain video action recognition using generative diffusion models.
Modified CompVis’ Stable Diffusion 1.5 and Hugging Face implementation of Modelscope (text-to-video model).
Injected code to extract cross-attention maps from UNet and gained deep understanding of generative diffusion models.
Injected into CompVis’ Stable Diffusion 1.5 and Hugging Face implementation of Modelscope.
Implemented multiple classification heads (MLP, 3D CNN, Transformer) for action classification on diffusion features.

PyTorch Hugging Face WandB Machine Learning Computer Vision

Full Stack Software Developer Intern | Fathomd

Aug 2021 - Jun 2023

Developed educational video games for business schools like MIT Sloan, with a focus on the instructor dashboard.
Created React components and added routes for new pages (Meteor/Ironrouter) in frontend.
Developed new REST API endpoints and modified DTO/DAO layers for MongoDB in a Dropwizard Java backend.
Migrated the instructor dashboard from BlazeJS to React.
Implemented a “classes” feature allowing instructors to manage tailored game sessions for different classes.
Collaborated with a team using tools like Jira and GitLab, gaining experience in agile software development workflows.

React Meteor.js Java Dropwizard REST API Full-stack

Projects

Caltech Course LLM Agent

LLM with retrieval-augmented generation (scraped course catalog and 2 years of student feedback) that answers Caltech course questions.

Python Flask LLM RAG Web scraping Langchain Langgraph React Selenium Vite Typescript

Caltech Course Graph

Interactive graph visualization and prerequisite finder for Caltech courses.

Vite React Typescript Tailwind CSS

Tennis Ranker

Web app that tracks tennis match scores and automatically ranks players.

Node.js Meteor.js MongoDB React Tailwind CSS

Coursework

CS 2: Data Structures

Fundamental data structures in Java, such as graphs, tries, and more.

Data Structures Java

CS 21: Decidability and Tractability

Study of computational complexity theory, exploring the boundaries between tractable and intractable problems.

Complexity Theory Algorithms

CS 24: Computing Systems

Operating systems and low-level programming concepts.

C Assembly Operating Systems

CS 38: Algorithms

Advanced algorithmic techniques, such as greedy approach, divide and conquer, dynamic programming, linear programming.

Algorithms Graph Theory Time complexity

CS 156ab: Learning Systems

Theory, algorithms, and applications of automated learning.

Machine Learning Statistical Learning

CS 155: Machine Learning/Data Mining

Popular methods in machine learning and data mining, with an emphasis on developing a working understanding of how to apply these methods in practice

Machine Learning Python Neural Networks

CS 148: Large Language and Vision Models

Deep dive into the math and implementation of modern large language and vision models.

Deep Learning PyTorch Large Language Models Computer Vision

CS 159: Advanced Topics in ML

Advanced machine learning concepts with focus on deep learning architectures and applications.

Deep Learning PyTorch Computer Vision

CS 172: Distributed Computing

Programming distributed systems, with applications in sensor networks, cloud computing, and machine learning/statistics.

Distributed Systems Distributed Algorithms

ACM 104: Applied Linear Algebra

Applications of linear algebra in computer science and engineering.

Linear Algebra Matlab

Ma 121a: Combinatorial Analysis

A survey of modern combinatorial mathematics, including graph theory and extremal problems.

Graph Theory Algorithms Combinatorics

Awards

William Lowell Putnam Mathematical Competition

Top 1000 | December 2023

Score of 19 (top 1000) in the largest North American intercollegiate math competition.

United States of America Computing Olympiad

Gold Division | February 2021

Gold Division in high school programming competition (theoretical CS, data structures, and algorithms).

American Math Competition/American Invitational Math Exam

AIME Qualifier | November 2021

Qualified for AIME 3 times by placing top 2.5% in American Math Competition (high school math competition), highest score of 6 on AIME.