Independent R&D / Manufacturing × AI
I build practical automation for manufacturing operations and knowledge work, connecting AI models to real workflows. I focus on reliable long-running execution through persisted state, bounded retries, validation, and recovery.
제조 운영과 지식 업무의 반복 작업을 자동화하고, AI 모델을 실제 업무 흐름에 연결합니다. 상태 저장, 제한된 재시도, 검증과 복구를 통해 장시간 실행의 안정성을 높입니다.
Project log
Built tools, active R&D, and experiments—clearly labeled by what I implemented and what remains exploratory.
Manufacturing workflows
Manual BOM and budget work repeats across drawings and spreadsheets. I built desktop tools to extract drawing BOMs, transform budget sheets, and organize project documents. BOM extraction is rule-based.
Built tools
Drawing utilities
Drawing conversion and updates involve repetitive desktop steps. I built Windows tools for BOM extraction and origin-coordinate changes. LibreDWG performs DWG-to-DXF conversion; the surrounding workflow tools are mine.
Built tools
Mail migration
Moving mail between accounts requires checkpoints and duplicate protection. I built a Python bridge for my Naver-to-Gmail migration with state tracking, duplicate detection, and post-import validation. It remains a single-account tool, not a public service.
Built tool
Agent control
Long-running tasks can lose state or repeat incomplete work. I’m building an orchestration layer with persisted progress, bounded retries, explicit stop conditions, and response validation. External model and browser platforms are integrations; the orchestration is mine.
Active development
Messenger integrations
Messenger-based knowledge capture needs provenance and stable revisions. I built the ingestion, provenance, and revision-handling layer, and tested constrained agent handoffs. It uses separate read-only connector clients; the clients and messaging platforms are external.
Integration experiment
AI memory
AI memory needs source context and user control. I built a conversation-wiki prototype with source provenance and revision history. Shared memory and persona continuity across AI tools remain a direction I’m exploring, not a shipped feature.
Prototype / exploration
I build the workflow layer; models, libraries, and platforms are integrations.
About
I build at the intersection of manufacturing operations and AI-enabled knowledge work.
My work focuses on recurring tasks: structuring engineering documents, moving information between systems, and making long-running workflows resumable and verifiable.
I use frontier and local AI models as components within automation systems, focusing on orchestration, validation, and reliable execution.
I’m exploring how these systems can evolve into reusable tools for manufacturing and knowledge work. Some are working utilities; others remain prototypes or experiments.
Contact
Manufacturing systems
AI automation
Agent workflows