About
Margrop
AI Infrastructure Engineer
I build and operate systems where AI agents, cloud platforms, self-hosting, and production engineering meet.
What I care about most is whether a system actually runs, stays stable, and can be diagnosed and repaired when it breaks. This blog therefore focuses on real deployments, migrations, incident reviews, automation, and postmortems — not just tutorials that look good on paper.
What I focus on
- AI Agent: coding agents, MCP, shared memory, model gateways, and agent workflows
- Self-hosting: Docker, NAS, Proxmox VE, home lab, and network infrastructure
- Engineering: Linux, Java, Go, databases, performance, observability, and production incidents
- Cloud Platform: service deployment, automation, permissions, token management, and production practices
These areas are not isolated. My goal is to embed AI capability into real engineering systems: with data, permissions, monitoring, and clear failure boundaries and rollback paths.
What my posts usually cover
- Context and background: where the problem appeared and what the original goal was.
- Verification process: what commands were actually run, what was observed, and which attempts failed.
- Root cause and fix: why it broke and what was changed.
- Acceptance criteria: how to prove the service recovered, the config took effect, or the deployment really went live.
Sensitive information about production environments, machines, or internal systems is always redacted.
Recommended reading
- First visit: see the full topic map at Start Here.
- Selected work for job applications: see the Portfolio.
- Latest field notes: browse the archive.
- Search by technology: browse tags.
- Code and automation: visit GitHub.
This blog is primarily in Chinese, with English versions for articles that have long-term reference value. Comments and corrections are welcome.