South Korea
Research Engineer working across multimodal AI, video intelligence, and generative models. Over 10 years, I have combined applied research with hands-on engineering to advance AI models and translate research into real-world systems across media, commerce, and industrial domains. Core experience • Multimodal AI: Foundation models, Video-LLMs, multimodal understanding, and model evaluation • Large-Scale Model Research: Training, optimization, and distributed experimentation with 10B–100B+ models • Generative Vision: Diffusion models, GANs, and AI systems for OTT and VFX • Applied Research Leadership: Leading cross-functional teams and bridging research with real-world applications Published in Pattern Recognition and WACV, with additional work at CVPR workshops. Named inventor on 14 AI patents.
- Led development of ML-based video-to-poster pipeline utilizing face detection & super-resolution models, significantly accelerating manual thumbnail creation process and increasing user engagement through character-focused thumbnails - Developed an advanced automated video QC system employing multimodal retrieval approach with diverse Vision ML models that detects 15+ quality parameters, dramatically reducing manual review time while implementing a fully automated AWS pipeline (Batch, Lambda, SQS) - Built ML-based video processing batch serving system on AWS infrastructure, designed for scalable and efficient content transformation workflows
- Led a cross-functional team of 9 AI researchers and engineers, increasing project delivery efficiency - Developed state-of-the-art generative models for face-swapping and re-aging, achieving 40% better visual quality than previous solutions - Published 2 research papers at renowned conferences, plus 2 arXiv preprints with one gaining notable attention in the field - Secured 10 patents for generative model applications - Architected and deployed scalable microservices on AWS
- Invented GAN-based battery anomaly detection system predicting automotive battery risks - Built CNN-based remaining life estimation model, winning $11K R&D Award for innovative approach - Created battery parameter estimation pipeline using AWS EC2, processing 10M+ simulation time series - Secured 5 patents for battery AI applications, including novel methods for thermal runaway prediction