seongwoonjo.
AI RESEARCH & ENGINEERINGUlsan, South Korea

Seongwoon Jo.

Thinking deeply.
Building efficiently.

I explore how to make AI more efficient — from the kernels that run it to the models that understand our world.

Explore my workGet in touch
RESEARCH INTERESTS
AI AccelerationOn-device AIEfficient TrainingLLMsWorld Models
01

A little about me

At the intersection of
models and systems.

My background spans computer science and industrial engineering at UNIST. I’m interested in the whole path from training a model to making it run well, with experience in research labs and AI startups.

ACADEMIC BACKGROUNDUNIST
MAJOR / MINORComputer Science and Engineering / Industrial Engineering
02

Selected work

From computation to understanding.

01SYSTEMS

Making LLMs run better

Analyzed operation-level latency in dynamic quantization and implemented a Triton kernel for LayerNorm fusion.

TritonLLM InferenceQuantization
Integrated Circuits and Systems Lab · 2025
02MODELS

Learning to simulate worlds

Trained Nano-GPT-based Transformers from scratch, adapted custom KV caching, and built a Valorant-domain world simulation model in collaboration with ETRI.

TransformersKV CacheWorld Models
ARI.AI · 2025
03APPLICATIONS

Finding signals in language

Collected and processed news and travel data, then fine-tuned language models for sentiment analysis in finance and travel.

KB-AIBERTKoBERTKoGPT
Financial Engineering Lab / Trip Builder · 2022–23
03

Where I’ve worked

Feb — Nov 2025INDUSTRY

ARI.AI

AI Researcher

Transformer training, custom KV caching, and world simulation for the Valorant game domain in collaboration with ETRI.

Advisor: Prof. Jooyeon Kim
Jul — Sep 2025RESEARCH

Integrated Circuits and Systems Lab

Undergraduate Research Intern · UNIST

Operation-level latency analysis for LLM dynamic quantization and Triton-based LayerNorm kernel fusion.

Advisor: Prof. Jongeun Lee
Jul 2022 — Mar 2023INDUSTRY

Trip Builder

Data Engineer & AI Researcher

Travel data crawling and processing, with KoBERT and KoGPT fine-tuning for sentiment analysis.

Apr — Dec 2022RESEARCH

Financial Engineering Lab

AICP Program Collaborator · UNIST

News collection and preprocessing, and KB-AIBERT fine-tuning for financial news sentiment analysis.

Advisor: Prof. Yongjae Lee
04

Education & recognition

EDUCATION

UNIST

Ulsan National Institute of Science and Technology

B.S. studies in Computer Science and Engineering
Minor in Industrial Engineering · Started March 2021

HONORS & AWARDS
2022

Gold Prize · 2nd Place

2nd KAIST–POSTECH–UNIST Data Science Competition

2021

Top Prize · Ministry of Science and ICT

U-Challenge Festival 2021

2021

Participation Prize

1st UNIST–POSTECH–KAIST Data Science Competition

05

Tools & teaching

MY TOOLKIT

Languages & web

C · C++ · Python · R
HTML / CSS / JavaScript · React

ML frameworks & tools

PyTorch · TensorFlow · Scikit-learn
JAX · Triton

PASSING IT ON

Learning through teaching.

  • Introduction to AI Programming I · UNISTTutor · Oct–Dec 2022
  • General Physics I · UNISTTutor · Apr–Jul 2022
  • Python for middle school students · Ulju The Dream CampusTutor · May–Dec 2022
  • Midam Scholarship Foundation & Aramco Korea Ulju-gun Education ProgramInstructor · 2021–2022
LET’S CONNECT

Good ideas start
with a conversation.

For research conversations, collaboration, or simply to say hello.