Work Experience Summary

3+ years of experience designing, training, and optimizing up to 519B scale large language models using distributed training techniques, contributing to a Top-3 ranking in the Korean World Best LLM project.
3+ years of experience partnering with cross-functional teams and clients to deliver AI solutions—including LLM fine-tuning, dataset synthesis, and evaluation system development—for projects valued at over $4 million.
3+ years of research experience in autonomous robotics and reinforcement learning, delivering 8 peer-reviewed publications and 2 patents through innovative solutions to real-world challenges. </aside>

WORK EXPERIENCE

ESFT Pipeline: Designed and implement ESFT (Expert-Specialized Fine-Tuning for MoE) in AWS SageMaker

Open-source Release: Main contributor to the development and release of SKT’s LLMs A.X-3.1, A.X-4.0, and A.X-K1 on Hugging Face, contributing to a Top-3 ranking in the Korean World Best LLM project
Large-Scale Distributed Training: Implemented multi-node training for 70B-scale Dense LLMs using FSDP, and 519B-scale MoE LLMs using Megatron-LM
Model Performance Improvements: Implemented Knowledge Distillation, improving general performance of 7B-scale LLM by 20%
Training Optimization: Developed "Preference Packing" technique for DPO that reduced training time and memory usage by 35%
Training Workflow Management: Designed a post-training workflow to resolve issues in use cases and improve collaboration with the data generation team
Evaluation Systems: Led centralizing SKT LLM Leaderboard project for efficient cooperation

Hynix HR chatbot Project (2024)
- Client: SK Hynix (over $2M)
- Synthesized data, fine-tuned LLM, for improving RAG performance over GPT-4o in HR domain without a decrease in general LLM performance
- Collaborated with the SK Hynix product team to design and implement the service pipeline and evaluation system
Coding assistant Project (2025)
- Client: SK AX (over $2M)
- Synthesized data, fine-tuned sLLM for inference speed, knowledge distilled from 70B-scale model for improving coding performance