Manufacturing Vision AI Engineer (PM/Lead) focused on taking deep learning projects from data acquisition to line deployment.
- Name: Jinwoo Jeong
- Role: AI Engineer, PM + Lead (Manufacturing Vision AI)
- Engineering Focus: Deep Learning-based Vision Inspection, Model Deployment, Inference Optimization, Cross-functional Project Delivery
- Tech Stack (Main): Python, C#, C++, PyTorch, ONNX, TensorRT
- Language Stack: Korean (Native), English (Work Communication), Japanese (Daily Communication)
- Email: jinwoo327@gmail.com
- GitHub: JuicyJeong
- Led end-to-end manufacturing AI projects: image acquisition, labeling, dataset setup, training/deployment, and on-site application.
- Covered practical tasks across Detection / Segmentation / Classification / OCR.
- Refactored Deepculus architecture from Boost.Python-style coupling to C# ↔ Python socket-based runtime communication.
- Improved deployment robustness with protocol design + packaging/security workflow (PyInstaller, PyArmor, archive protection).
- Applied model optimization workflow with ONNX / TensorRT and large-image tiling-based inference handling.
- Collaborated cross-functionally with optics/system teams for visibility checks, tact-time alignment, and customer-specific integration.
- PyTorch, Detection, Segmentation, Classification, OCR
- Serving/Deployment: ONNX, TensorRT
- LLM Workflow: Prompt Engineering, Agentic Workflow, MCP, vLLM
- Programming: Python, C#, C++, Bash/Shell
- OS/Infra: Ubuntu, macOS, Windows, NAS (Synology / TrueNAS)
- Environment: Conda, uv, Git/GitHub, SSH
- ACTOR retraining for avatar motion generation
- Retrained a public motion generation model with modified datasets and visualized outputs in Unity3D.
- Motion similarity automation tool (TMR-based)
- Automated embedding generation/similarity evaluation and improved throughput with custom loader + parallel processing.
- Unity3D + IMU real-time motion visualization
- Connected IMU sensor data to SMPL-X avatar joint rotations with calibration/alignment scripts.
- Event booth crawling & visualization system
- Crawled exhibitor data and organized/visualized it in Google Sheets for planning.
- Event account info collection pipeline
- Automated account collection and structured shared operation data.
- Music chart collection & auto-posting bot
- Operated a real-time chart data pipeline with automated posting workflow.
- VRST 2024 (Poster)
- ICDDS 2024 (Oral)
- Korean Institute of Industrial Engineers (2022)
- Korean Software Conference (2023, 2024)
- ADsP (Advanced Data Analytics Semi-Professional)
- TOEIC Speaking IH
- Email: jinwoo327@gmail.com
- GitHub: JuicyJeong