Curriculum vitae Last updated — October 2026
Download as PDF
Education M.S., Interdisciplinary Program in Artificial Intelligence Sep 2025 – Aug 2027
Seoul National University
GPA: 4.20 / 4.30 (in progress) B.S. (Double Major), School of Computing × Mathematical Sciences Feb 2017 – Aug 2025
Korea Advanced Institute of Science and Technology
Semiminor in Artificial Intelligence. On leave 2020 – 2024 — industry R&D at Humelo, Inc., including two years of alternative military service.
Publications Efficient probabilistic weather forecasting at quarter-degree resolution with two-sided transport attention 2026
Junu Choi , Youngjoon Hong
Preprint · arXiv:2610.35545 Posterior Transport Matching for Inverse Problems 2026
Junu Choi , Youngjoon Hong
Experience Graduate Researcher (M.S.) Sep 2025 – present
Seoul National University
Advisor: Prof. Youngjoon Hong. Astra: Kantorovich Attention (two-sided transport attention) for efficient quarter-degree probabilistic weather forecasting — arXiv:2610.35545. Posterior Transport Matching for Inverse Problems: minimum-energy transport matching for posterior sampling — under review, ICLR 2027. Undergraduate Researcher Jan – Mar 2025
Visual AI Group, KAIST
Investigated population-simulated resampling to expand the design space of sequential Monte Carlo samplers. Undergraduate Researcher Mar – Aug 2024
MLILAB, KAIST
Fine-tuned conversational LLMs on medical instruction Q&A using parameter-efficient adaptation. Developed an evaluation protocol to address GPT-4 evaluator bias and scoring variability. AI Researcher (Speech & Multimodal) Sep 2020 – Jan 2024
Humelo, Inc. — incl. two years of alternative military service (industrial technical personnel)
Production text-to-speech and voice-conversion systems across multiple product lines. Proposed Style Contrastive Adversarial Neutralization (SCAN) for vocal-style disentanglement, combining bi-adversarial training with contrastive feature compaction. Implemented an LTU-2-family audio–text understanding system (Llama × Whisper-AT) with partial LoRA / QLoRA training. Stabilised auto-regressive TTS via Lipschitz-preserving discriminators and 3-step causal training. Research Intern Jul – Sep 2020
Humelo, Inc.
GMM-based emotional control of speaker embeddings; multi-band mel-GAN vocoder.
Selected projects LightTransporter — diffusion-based image relighting Fall 2024
KAIST · CS492D: Diffusion Models and Their Applications
Led a three-person project using a Light-Image Encoder to condition diffusion on source/target environment maps and camera pose; evaluated on OpenIllumination. Feature-level speculative decoding Mar – Jun 2024
KAIST · Team project
Led a team project combining intermediate-feature speculation, knowledge distillation, and sparse rejection for speculative decoding.
Honors & awards AI Fellowship (AI Graduate School Support Program) 2025 – present
Seoul National University, Interdisciplinary Program in Artificial Intelligence
Government-sponsored graduate fellowship providing full tuition waiver and monthly stipend. National Science & Technology Scholarship (KAIST Scholarship) 2017 – 2025
Korea Advanced Institute of Science and Technology
Government-sponsored full tuition waiver and monthly stipend for 8 semesters (excluding leave). Qualcomm Innovation Award 2019
KAIST × Qualcomm AI Hackathon