Summary
Overview
Work History
Education
Skills
Publications
Accomplishments
Timeline
Generic

Soonwon Ka

ML Research Engineer
Seoul,South Korea

Summary

Innovative Tech Lead at NAVER Corp with expertise in LLM frameworks and a strong focus on performance optimization. Successfully enhanced AI solutions for search/feed experiences and contributed to generative AI services. Proven ability in technical problem solving and communication skills, driving impactful research outcomes and productivity improvements. Personally a global nomad into local culture exploration.

Overview

9
9
years of professional experience
8
8
years of post-secondary education
3
3
Languages

Work History

Tech Lead, ML Research Engineer

NAVER Corp
Seongnam-si, Gyeonggi-do
11.2023 - Current
  • Collaborated with product managers to define technical requirements and develop optimized solutions for AI transformation of feed & search experiences, i.e., Homefeed & AI Briefing.
  • Enhanced inference efficiency (QPS per GPU) by 20 times without generation quality loss by LLM distillation.
  • Conducted multimodal model research for image-to-text alignment (Retrievers & VLMs).
  • Played a role as a solver to proactively identify and remove bottlenecks in progress.

ML Research Engineer

NAVER Corp
Seongnam-si, Gyeonggi-do
04.2021 - 10.2023
  • Conducted LLM-based research & development to release cutting-edge generative-AI services for both sellers and customers, i.e., AI Cuesheet and CLOVA for Writing (RAG).
  • Presented research findings at conferences and workshops, contributing valuable insights to the broader scientific community (from model pre-training to AI writer application).
  • Implemented meta outputs of generative AI pipelines to improve team productivity (LLM-based synthetic data generation & filtering).

Researcher Assistant & Engineer

Soundl.ly
01.2016 - 04.2019
  • Implemented and optimized a software based ultrasound communication modem for sound QR marketing solutions among smartphones.
  • Conducted rigorous quality assurance tests, identifying areas of improvement for product optimization.

Education

Master of Science - Electrical And Computer Engineering

Seoul National University
Seoul, South Korea
03.2014 - 02.2016

Bachelor of Science - Electric Engineering

Seoul National University
Seoul, South Korea
03.2008 - 02.2014

Skills

Technical problem solving

Team collaboration

Performance optimization

Data-driven decision making

LLM framework knowledge (HuggingFace, VLLM, SGLANG)

Publications

  • Lee et al., Navigating the Path of Writing: Outline-guided Text Generation with Large Language Models, Feb. 2025 (accepted to NACCL 2025 Industry Track).
  • Cho et al., DSAI: Unbiased and Interpretable Latent Feature Extraction for Data-Centric AI, Dec. 2024 (Arxiv).
  • Yoo et al., HyperCLOVA X Technical Report, Apr. 2024 (Arxiv).
  • Park et al., HyperT5: Towards Compute-Efficient Korean Language Modeling, Jul. 2023 (ACL 2023 Industry Track).
  • Kwak et al., No Entry: Anti-noise Energy Detector for Chirp-based Acoustic Communication, Jun. 2019 (SECON 2019).
  • Ka et al., Near-ultrasound Communication for TV's 2nd Screen Services, Oct. 2016 (MobiCom 2016).

Accomplishments

  • Oral presentation at NAVER DAN 2024 for LLM distillation (https://dan.naver.com/24/sessions/634)
  • Online presentation at D2 NAVER 2024 for LLM distillation (https://d2.naver.com/helloworld/8866888)
  • Online presentation at D2 NAVER 2024 for writing AI (https://d2.naver.com/helloworld/8011540)
  • Oral presentation at NAVER DEVIEW 2023 for seq2seq model domain adaptation (https://deview.kr/2023/sessions/554)

Timeline

Tech Lead, ML Research Engineer

NAVER Corp
11.2023 - Current

ML Research Engineer

NAVER Corp
04.2021 - 10.2023

Researcher Assistant & Engineer

Soundl.ly
01.2016 - 04.2019

Master of Science - Electrical And Computer Engineering

Seoul National University
03.2014 - 02.2016

Bachelor of Science - Electric Engineering

Seoul National University
03.2008 - 02.2014
Soonwon KaML Research Engineer