John Kim

Computer Science student at the University of Virginia. I work on machine learning research, build full-stack applications, and compete in programming competitions.

ResumeContact

About

I am a researcher and developer driven by the challenge of aligning artificial intelligence with human values. At ML@UVA, I collaborate with Johns Hopkins APL to study decision-making in LLM agents, exploring how we can make AI systems more reliable and interpretable. Beyond research, I enjoy building practical full-stack applications and solving challenging math & cs problems.

Education

University of Virginia

August 2025 - Present

B.S. in Computer Science

GPA: 4.0 / 4.0

Thomas Jefferson High School for Science and Technology

August 2021 - June 2025

Advanced Studies Diploma

GPA: 4.463 / 4.0

Experience

ML Researcher

ML@UVA × Johns Hopkins APL

September 2025 - Present
  • • Explored the strategy game Diplomacy to study human-aligned decision-making in LLM agents
  • • Investigated methods for aligning AI decisions with human behavior distributions using fine-tuning
  • • Experimented with steering vectors and prompting techniques to mimic distinct human playstyles

Full Stack Developer Intern

MySmaX Lab (AIoT Startup), Seoul National University

June 2025 - August 2025
  • • Built and deployed an AI agent to automate IoT workflows using Model Context Protocol (MCP)
  • • Applied machine learning–based anomaly detection to analyze IoT device data
  • • Served as a primary contributor to MySmaX’s user-facing production website using Next.js

Competitive Programming & Mathematics

Clubs & Competitions

2021 - Present
  • • USACO Gold Division
  • • American Invitational Mathematics Examination (AIME) Qualifier (4 times)
  • • Member of the Putnam and ICPC Club at UVA; Former member of TJ Varsity Math Team
  • • Jane Street Estimathon Winner in 2024, 2025

Skills

Programming Languages

Python
Java
C++
JavaScript
TypeScript
HTML
CSS

Web Development

React
Next.js
Django
FastAPI
Tailwind CSS

Developer Tools

Git
Docker
Linux Shell
Jira
AWS
CI/CD

AI & ML

PyTorch
TensorFlow
LangChain
Agno
MCP
Unsloth
YOLO

Projects

Hoos' Plan

Full-stack UVA course planning website that helps students build semester-by-semester academic plans and validate enrollment requirements including prerequisites, corequisites, and other restrictions.

TypeScript
Next.js
Python
Prisma
Tailwind CSS
  • • Built full-stack from scratch — designed the database schema, REST API, and React frontend
  • • Scraped and parsed UVA's course catalog to populate prerequisite and corequisite graphs
  • • Validates complex enrollment chains in real-time as students drag courses into semesters

ML@UVA Club Website

Ongoing development and maintenance of ML@UVA's official organization website.

TypeScript
Next.js
React
Tailwind CSS
  • • Migrated the codebase from Vite to Next.js to secure API keys server-side
  • • Completely redesigned the UI across all pages
  • • Added and reorganized site pages and kept content up to date

Alpine Ski Racing AI Analysis Model

Deep neural network using CNNs to provide ski racers with quantitative feedback from video analysis.

Python
YOLO
PyTorch
CNNs
  • • Built a deep neural network using CNNs to analyze ski racing technique
  • • Used YOLO for pose extraction and frame-by-frame analysis
  • • Achieved results that consistently aligned with real race performances

Offline AI Model for North Korea

Offline generative AI solution for distributing reliable information in North Korea. Received $7,500 grant from Human Rights Foundation.

Python
LangChain
HuggingFace
Unsloth
  • • Developed offline AI solution by fine-tuning LLMs for edge devices
  • • Received $7,500 grant from Human Rights Foundation
  • • Optimized for resource-constrained hardware

Contact

I'm always open to discussing new projects, opportunities, or just having a chat about technology and AI.