Operational AI and infrastructure engineer building reviewable agent, data, manufacturing, and secure-workflow systems.

Live systems gallery · LinkedIn · Private collaboration route

I build systems whose control flow, failure modes, data boundaries, and verification path can be inspected. Public demos use synthetic or fixture data unless a repository explicitly states otherwise; planning documents do not establish customer deployment or revenue.

Review the curated architecture index, repository evidence map, and quality gate before treating a demo or planning document as production evidence.

1. fab-ops-yield-control-tower: inspect staged fab/scanner operations, release gates, and the documented MES/SCADA boundary.
2. AegisOps: run the deterministic replay suite and inspect schema, rubric, incident-report, and operator-handoff behavior.
3. enterprise-llm-adoption-kit: review RBAC, redaction, injection checks, evaluation gates, and environment-gated providers.
4. stage-pilot: reproduce the 60-case, 30-mode parser-recovery benchmark and inspect the 33.3% / 66.7% / 90.0% strategy results.
5. lakehouse-contract-lab: run the contract, quality, deterministic-artifact, and API verification path.

For a visual first pass, open the live systems gallery and then use each repository's own verification command.

1. Open the live gallery and choose one operational system, one runtime, and one data system.
2. Read that repository's System Overview, operating boundary, architecture notes, and quality gate.
3. Run the documented local verification command before relying on benchmark or readiness claims.
4. Use the account-level cloud and AI architecture blueprint, machine-readable manifest, and architecture validator for cross-system context.

For a lane-oriented technical review, use this order:

1. fab-ops-yield-control-tower
2. AegisOps
3. enterprise-llm-adoption-kit
4. stage-pilot
5. lakehouse-contract-lab
6. aix-pilot

The current public inventory contains 30 active public original repositories and 15 archived public original repositories. The dated 35-repository publication and commercial catalog remains a historical snapshot; memoryflow-lab is represented in this active public inventory as systems-performance research evidence rather than a speculative service SKU.

• AI assistance is used for implementation drafts, tests, documentation, and review; ownership is demonstrated through requirements, debugging, verification, and explainable decisions.
• Synthetic results are not presented as customer deployments or universal product performance.
• Hardware measurements retain environment and protocol metadata and state where they do not generalize.
• Private case studies expose sanitized public demos rather than inaccessible source or architecture links.
• Repositories fail explicitly when required inputs, credentials, or operating assumptions are absent.

• Computer Science coursework, Korea National Open University, 2026–present.
• Computer Science degree candidate through the Bachelor's Degree Examination for Self-Education, expected November 2027.
• Microsoft AI School, 8th cohort.
• AI Semiconductor Architecture Design and Performance Optimization, Seoul ICT Innovation Square / KAIT, completed July 2026.
• IT infrastructure operations: monitoring, access administration, backup, incident response, workspace administration, vendor coordination, and shift handoff.
• Six-person microwave (MW) communications squad leadership, ROK Defense Communication Command.
• Microsoft Azure AI Fundamentals (AI-900).

For professional collaboration or a scoped technical discussion:

• LinkedIn
• Private portfolio route — falls back to LinkedIn unless the dynamic endpoint is verified
• GitHub