JEONG JINWOO

Manufacturing Vision AI PM/Lead and AI Engineer

jinwoo327@gmail.com

Github

010-3277-6270

Date of Birth: 1995.03.27

PROFESSIONAL SUMMARY

Profile photo of Jeong Jinwoo
  • I have hands-on experience across the full project cycle in deep learning-based vision inspection: image acquisition, defect criteria definition, labeling, model training/deployment, performance improvement, and equipment-level rollout across multiple customer projects.
  • I translate priorities into executable plans, making practical decisions on what matters most while balancing schedule, quality, and risk.
  • I collaborate closely with teammates to align strengths and challenging tasks, continuously improving team execution efficiency.
  • I actively track new technologies, then validate and experiment from an adoption perspective to ensure real business applicability.
  • Beyond vision AI, I am strongly interested in workflow automation, data crawling, 3D pose estimation, and motion generation tasks.

CORE SKILLS

AI Engineering

PyTorch, Detection/Segmentation/Classification, ONNX, TensorRT, Prompt Enginnering, Agentic Workflow, MCP, Local LLM

Engineering Stack

Programming: Python, C#, C++, Bash/Shell

OS/Infra: Linux(Ubuntu), Windows, NAS(Synology, TrueNas Scale)

Environment: Conda, uv, Git/GitHub, SSH

Certifications

ADsP (Advanced Data Analytics Semi-Professional)

TOEIC Speaking IH

Driver's License Class 1 (Korea)

Language

English: TOEIC Speaking IH, comfortable with daily conversation and practical work communication

Japanese: Listening/Speaking - able to communicate in daily conversations without difficulty

WORK EXPERIENCE

HiveVision Co., Ltd. | AI/System Development Team

AI Engineer, PM + Lead | 2025.05 - Present

Manufacturing Vision AI Project Lead

  • Led manufacturing vision projects end-to-end as PM and AI engineer, covering data acquisition, labeling, dataset setup, and training schedule management.
  • Defined technical direction and proposals by analyzing customer requirements across detection, segmentation, classification, and OCR tasks.

Deepculus Architecture Refactor

  • Refactored a tightly coupled Boost.Python-style structure into socket communication between the C# application and Python runtime.
  • Designed message protocols and applied Python EXE packaging (PyInstaller), source obfuscation (PyArmor), and archive protection for stronger deployment stability and security.

Model Benchmarking and Adoption

  • Applied partially tuned open-source backbones (DINO, SAM) to deep learning models.
  • Planned next-pipeline onboarding of a new YOLO model and validated the speed-accuracy balance.

Inference Optimization

  • Used tiling-based inference for large images and tuned tile batch size to reduce tact time versus single-batch inference.
  • Converted trained models to ONNX and TensorRT and deployed them in customer solutions.

Cross-Functional Collaboration

  • Collaborated closely with optics and system development teams for image acquisition and on-site equipment integration.
  • Proactively responded by understanding each team’s scope and required effort in advance.

PROJECTS

Graduate Projects

Research on Human Motion-related topics

Model Retraining for Avatar Motion Generation with Modified Dataset

  • Keyword: Motion Generation, Unity3D, Dataset Adaptation, SMPL-X, 3D Pose
  • Retrained the public motion-generation model (ACTOR) on specific motion data to generate new motion sequences.
  • Built a batch conversion script for over 10,000 motion samples and visualized outputs in Unity3D.

Motion Similarity-based Classification Automation and Visualization Tool

  • Keyword: Motion Similarity, Vector Embedding, Automation, Visualization
  • Automated TMR encoder-based embedding generation and similarity comparison to reduce qualitative review time.
  • Improved comparison throughput by implementing a custom data loader and parallel processing.

3D Motion Visualization by Integrating Sensor Data into a Unity3D Avatar

  • Keyword: Unity3D, C#, SMPL-X, IMU, Pose Estimation, Real-Time Motion Capture
  • Connected IMU sensor data in real time to joint rotations of a Unity3D SMPL-X avatar.
  • Implemented sensor calibration and coordinate alignment scripts for stable demonstrations.

Personal Projects

Projects focused on data collection and automation to reduce repetitive work and improve operational decision-making.

Comic Event Booth Crawling and Visualization System

  • Keyword: Web Crawling, BeautifulSoup, Pandas, Google Sheets API
  • Crawled event booth information and visualized it in Google Sheets for operations decision-making.
  • Improved booth planning accuracy through automatic updates and tag filtering.

Event Booth Participant Account Collection and Structuring System

  • Keyword: Web Crawling, Bot, Selenium
  • Automated participant account collection, structured data in sheets, and standardized team-sharing workflows.
  • Linked key tag statistics with account data to support pre-event operations planning.

Music Chart Data Collection and Twitter Auto-Posting System

  • Keyword: Web Crawling, BeautifulSoup, Tweepy, Twitter(X) API
  • Structured real-time chart data and operated an automated posting pipeline using Twitter API.
  • Automated daily and weekly update flows to reduce repetitive work and improve operational continuity.

EDUCATION

Chung-Ang University, Graduate School of Advanced Imaging Science, Multimedia & Film (AI Imaging) 2023.03 - 2025.02

Kyonggi University, Industrial Management Engineering 2015.03 - 2023.02

Songlim High School 2011.03 - 2014.02