Congqi Huang

AI Engineer Intern @ China Railway Major Bridge Engineering Group Co., Ltd

About

I build practical AI systems and developer tools that connect research with real-world workflows. I am also an open-source contributor to OpenClaw and PicoClaw, where I work on agent runtime, integrations, configuration, and developer experience improvements. During my AI Engineer internship at China Railway Major Bridge Engineering Group, I developed UAV-based inspection systems for construction safety and bridge damage detection. I worked on YOLOv9c, RT-DETR, OpenCV, segmentation pipelines, and PyQt5 tooling, improving large-object detection from 30% to 87%, small-object detection from 0% to 91%, and reducing inspection time from half a day to about 20 minutes. Outside internships, I created ContextForge, a local-first repository-to-agent context CLI for Codex, Claude Code, and Cursor, and I write at wuster.store about AI agents, Go backend engineering, computer vision, and software systems. I am especially interested in AI engineering, computer vision, agent systems, developer tooling, and end-to-end product building. Core stack: Python, Go, TypeScript, Java, OpenCV, YOLO, RT-DETR, PyQt5, Vue, and Spring Boot.

Country

United States

City

Chicago

Industry

Computer Software

Skill

Computer Vision, Object Detection, Image Segmentation, YOLO, OpenCV, Python (Programming Language), Machine Learning, PyQt5, Deep Learning

Experience

AI Engineer Intern

China Railway Major Bridge Engineering Group Co., Ltd

2023-9 - 2024-5 · 9 mos

Wuhan, Hubei, China

Responsible for developing UAV-based intelligent inspection systems for construction site safety and bridge damage detection. A UAV-based Inspection System for Construction Scenarios: • Developed and optimized YOLOv9c and RT-DETR detection algorithms to identify both large objects (e.g., vehicles, cranes) and small objects (e.g., people, helmets). • Trained models on a dataset of 6,007 images; improved detection accuracy for large objects from 30% to 87% and for small objects from 0 to 91% through cross-validation and transfer learning. • Implemented real-time object detection and automatic alarm functions using OpenCV, enhancing manual inspection efficiency to under 5 minutes. A UAV-based Bridge Damage Detection System: • Applied YOLO models and segmentation algorithms to automatically detect cracks and rust, calculating maximum crack widths to millimeter-level accuracy using the inscribed circle method. • Designed a PyQt5-based user interface (UI) to support image uploads, visualization and annotation of damaged areas, and generation of detailed detection reports. • Reduced bridge inspection time from half a day to 20 minutes and significantly improved the efficiency and precision of inspecting hard-to-reach areas.