ML Engineer, specializing in Computer Vision and VLM (Vision-Language Models) multimodal learning, and deep learning. Proven expertise in translating cutting-edge research into scalable AI systems for real-world industrial applications, driving significant improvements in efficiency and accuracy across diverse domains. Adept at leading complex R&D projects, optimizing model performance for deployment at scale, and delivering data-driven solutions that directly impact operational outcomes.
Senior Computer Vision Research Engineer
Seoul, Seoul, Korea (Republic of)
→
Summary
Led the production rollout of the AI-based RoadVision platform and delivered client-facing R&D projects, significantly enhancing road infrastructure assessment through advanced computer vision.
Highlights
Directed the production rollout of the AI-based RoadVision platform across 13 provinces, successfully integrating computer vision pipelines for comprehensive road infrastructure assessment using PyTorch and Docker.
Developed and deployed real-time object detection models for road surface analysis, achieving 95% mAP accuracy, which informed data-driven asset management decisions and reduced inspection time by 80%.
Successfully delivered over 10 client-facing R&D projects in road infrastructure assessment, combining strong technical solutions with collaborative team leadership to meet critical client needs.
Optimized model inference for lightweight hardware using TensorRT, ONNX, and PyTorch, enabling efficient deployment at scale and ensuring sustainable, high-performance operations.
Computer Vision Engineer
Seoul, Seoul, Korea (Republic of)
→
Summary
Developed and deployed advanced computer vision models for damage segmentation and defect detection, optimizing real-time performance and integrating solutions with cloud services.
Highlights
Adapted and fine-tuned UNet architecture for construction site damage segmentation, achieving a 0.7 Dice loss and 0.9 accuracy, demonstrating advanced transfer learning expertise.
Developed and deployed segmentation and detection models with an mIOU of 0.8 for the CrackViewer.com platform, delivering a web-based solution integrated with AWS cloud services.
Designed a high-accuracy (99%) defect detection and classification algorithm for carbon material quality assessment, significantly streamlining quality control processes and reducing manual inspection efforts.
Engineered and optimized a real-time road damage detection system using ONNX and TensorRT, achieving a 4× FPS increase (100 FPS) for efficient large-scale highway monitoring.
Company
→
Summary
Developed an AI agent that converts natural language questions into SQL queries for analyzing road infrastructure defects. Built with LangChain and Claude AI, the system uses ReAct reasoning for multi-step query processing, features self-correcting SQL generation, and integrates with YOLO detection pipelines. Deployed on Streamlit Cloud with real-time analytics dashboard.
→
Summary
Constructed a vision-language segmentation pipeline by modifying CLIP (VIT-L/14) to perform dense prediction, integrating visual and textual features for improved crack segmentation accuracy using PyTorch.
→
Summary
Leveraged YOLOv8 to build a highly accurate (98%) defect detection system for real-time quality control in manufacturing.
→
Summary
Designed and implemented YOLOv3-based object detection modules using PyTorch to enable immediate road damage identification within the RoadVision system.
→
Summary
Built a comprehensive platform for concrete crack assessment, combining AI (Python) and a web framework (Django).
→
Summary
Developed a web platform with YOLOv5 (PyTorch) for precise pavement condition analysis, deployed using Django and Docker.
→
Summary
Addressed pavement image dataset limitations using a GAN (TensorFlow), achieving a low 0.1 loss in synthetic images.
→
Summary
Designed custom image processing algorithms to achieve 99% defect detection accuracy on the production line.
→
PhD and Master
Structural Engineering
Grade: 3.87/4.5
→
Bachelor
Mining
Grade: 3.82/4.0
Deep Learning (PyTorch), Object Detection, Image Segmentation.
VLMs, LLMs, RAG, APIs (openai, claude), Agent Workflows, Multi-Agent Systems.
Python, C++, Gradio, Django, OpenCV, Scikit-learn.
Model Deployment (Docker), CI/CD Integration, Model Monitoring, Data Pipelines.
Team Leadership, Project Management, Technical Communication, Problem-solving.
Issued By
DeepLearning.AI
Issued By
Coursera
Issued By
Udacity
Issued By
Udacity
Issued By
edX
Published by
REAAA
Summary
Developed a novel method for crack segmentation leveraging text-guided features within a vision-language model, enhancing precision for structural analysis.
Published by
Sensors
Summary
Submitted for publication, this work presents a comprehensive approach to automated pavement condition index assessment using deep learning and image analysis techniques.
Published by
Sensors
Summary
Authored a paper on a comprehensive deep learning and image analysis approach for automated pavement condition index assessment.
Published by
International Journal of Pavement Engineering
Summary
Researched and implemented region-based convolutional neural networks for automated detection of pavement distress, improving efficiency and accuracy in infrastructure monitoring.
