Available for research & ML-engineering collaborations
AI/ML Engineer · Computer-Vision Research
I build production LLM systems and publish first-author research on semi-supervised segmentation. I care about closing the gap between what works in a paper and what holds up at scale.
First-author work on learning dense predictions from very few labels. CW-BASS & FARCLUSS were the former #3 and #2 global SSSS SOTA with ResNet-101; PixCon is the current #1 on foundation-model features.
A clean-positive pixel memory bank that admits only labeled pixels, keeping contrastive positives free of pseudo-label noise on DINOv2-scale features. Takes the #1 spot in semi-supervised segmentation alongside UniMatch V2.
Fuzzy top-K pseudo-labeling, entropy-normalized uncertainty weighting, per-batch class rebalancing, and prototype contrastive regularization. Formerly #2 globally on Cityscapes & Pascal VOC (78.8% / 78.2% mIoU, ResNet-101).
Confidence-weighted cross-entropy with Sobel boundary regularization and bounded-sigmoid dynamic thresholding. Formerly #3 globally on Pascal VOC 2012 (77.15% mIoU) and top-10 on Papers With Code.
Open-source tools, research explorations, and full-stack products.
Read, fill & edit Korean HWP (Hancom Office) documents in Python — text extraction for LLM/RAG, programmatic form-filling, and corruption-free binary rewrite.
Stealthy backdoor triggers in CLIP-style models via dual-embedding alignment in the joint visual-textual representation space.
Targeted perturbations for multi-label classifiers that exploit semantic label co-occurrence dependencies.
Systematic evaluation of weak-to-strong consistency (UniMatch) and self-training (ST++) — the foundation for CW-BASS and FARCLUSS.
Depth estimation and point-cloud generation for autonomous driving, with an improved depth-to-pointcloud pipeline on KITTI.
Why declared context length overstates the effective one — analyzing positional-encoding bias and the STRING shifting method across Llama, Mistral & Qwen.
Real-time traffic monitoring with Three.js 3D visualization, edge-AI inference, and predictive analytics for smart-city integration.
AI-powered disease identification with severity scoring, structured learning modules, community data contributions, and downloadable datasets.
20+ subjects, 70+ lessons, progress tracking with study streaks, a learning-analytics dashboard, full auth, and a responsive UI.
Gractor · Seoul, Korea
Primary engineer building production LLM systems (RAG, tool-calling agents, evaluation harnesses) and computer-vision pipelines with edge inference, across live smart-city government platforms.
Korea University · Pattern Recognition & ML Lab
Advised by Prof. Seong-Whan Lee (IEEE Fellow). Three first-author segmentation papers ranked #1 / #2 / #3 globally; ~10K LOC of PyTorch multi-GPU training infrastructure; Korean patent filed (autonomous-driving perception).
GliT (GLITEC) · EdTech
Built offline-first mobile learning products reaching 500+ students and 80,000+ learning sessions, and led STEM-education initiatives.
Korea University · 2023 – 2026 · GPA 3.78 / 4.0
Global Korea Scholarship (sole Zimbabwe awardee) · BK21 Research Fellowship · Advisor: Prof. Seong-Whan Lee.
Korean National Assembly recognition
Global Korea Scholarship (full MSc funding) · BK21 Research Fellowship (2023–2026) · IJCNN 2025 first-author acceptance (Rome).
IBM Applied AI Professional Certificate
Modern Robotics Specialization — Northwestern
Semantic Segmentation with SageMaker — Amazon
ML Pipelines with Azure ML Studio — Microsoft
Foundations of Project Management — Google
AWS S3 Basics — Amazon Web Services
Neuroscience — Emory University
Game Development with Scratch — MIT