Ikjun Choi
EECS undergraduate at GIST
I build things to understand technology and its value.
I'm an EECS undergraduate at GIST, also studying business and economics. I enjoy exploring new technologies, building products, and understanding how they can make a difference in the real world.
Through software development and AI research, I've become increasingly interested in not just how technology works, but why it matters and where it can create value.
Recent
- Jun 2026AwardAI Engineer World's Fair Hackathon3rd place with Rote, a continual-learning computer automation platform.
- Jun 2026EducationUC Berkeley Summer SessionEight weeks in Berkeley for CS161 Computer Security and CS188 Introduction to AI.
- Mar 2026CareerJoined Real-Lab @ GISTUndergraduate research intern working on diffusion language models.
- Jan 2026ProjectGIST ChatbotStarted building a chatbot over GIST's official documents with GSA InfoTeam.
Selected work
All projectsDiffusion Language Models
ResearchMar 2026 – Present
Undergraduate research on more practical training and generation for diffusion-based language models.
- With
- Real-Lab @ GIST
- Role
- Undergraduate research intern
- PyTorch
- Hugging Face
- Python
SPRIT
ProductMar 2024 – Present
A mobile app that helps people build a steady reading habit.
- With
- Personal
- Role
- Product planning · App development
- Flutter
- Next.js
- NestJS
- Docker
GIST Chatbot
ProductJan 2026 – Present
A retrieval-based assistant that answers questions about GIST from the school's official documents.
- With
- GSA InfoTeam
- Role
- Answer generation system · Full-stack development
- React
- TypeScript
- NestJS
- MCP
Writing
All writing- How diffusion language models workInstead of emitting one token at a time, a diffusion language model restores a whole sentence. A walk through the objective and the sampling loop of masked diffusion LMs, with the equations.2026-10-05
- Automating stock research reports with a brokerage API and an LLMA pipeline that pulls price and financial data from the Korea Investment & Securities Open API, has the ChatGPT API analyse it, and posts the result to Slack. Data normalisation, prompt structure, pinning and validating the output format, cost control, and what I would change.2024-06-03
- I built a service alone. Will anyone like it?What I learned taking SPRIT, a reading-habit app, from idea to launch to its first month of operation on my own: how I found the first users, what the numbers said, and what building a one-person product is actually like.2024-04-02