Operations Support: track mind map
A mind map of the Operations Support track, the safeguarding layer that wraps this repository's AI Engineering course (modules M0–M33). It is a competency map / lens, not a new course and not a rename: AI engineering is the backbone (you build the system), and operations support is the layer that keeps what you built running, supported, and recoverable in production.
The course keeps its name and structure. This folder is entirely additive. It sits alongside the roadmap map in
../roadmap-mindmap/and the 34 module folders, all untouched. The track's hands-on modules are Part E (M31–M33); this map shows how every module across the course contributes to operating and supporting the system, plus the gaps still worth building.
How the two layers fit
| Layer | Question it answers | Where it lives |
|---|---|---|
| AI Engineering (backbone) | how do I build AI apps? | M0–M30 (Parts A–D) |
| Operations Support (safeguarding layer) | how do I run & support what I built, and recover when it breaks? | Part E (M31–M33), drawing on M18–M30 |
Operations support safeguards the whole stack: the architecture (uptime, incidents), the databases (the RAG index: freshness, backups, retention), and the builds (canary, rollback, secret rotation).
The files (use whichever your tool likes)
| File | Open it with |
|---|---|
ai-operations-support.xmind |
XMind: double-click to open directly. |
ai-operations-support.opml |
XMind, MindNode, Freeplane, OmniOutliner, most mind-map tools (File → Import → OPML). |
ai-operations-support.md |
Any editor (readable as-is); XMind also imports Markdown (File → Import → Markdown). |
coverage.md |
Auto-derived audit: every module M0–M33 → where it sits on this map, plus every [NEW] node. |
generate_mindmap.py |
The single source. Edit the OUTLINE, run python3 generate_mindmap.py, and all four files regenerate in sync. |
If the
.xmindever won't open in your XMind version, import the.opmlor.mdinstead, same content.
How to read the node tags
Every leaf is tagged so you can see where it is taught:
[M22],[M10 · M23]— taught in that existing module.[M31]/[M32]/[M33]— taught in the new Part E operations-support modules.[NEW]— an operations-support topic the course does not cover yet: a proposal on the map, not a built module. (There are 9 left; the rest of the original gaps are now Part E modules.)
The thirteen branches (the operational arc)
- The AI Operations Support role — what the safeguarding layer is, the lenses (LLMOps/AgentOps/AIOps).
- Foundations you operate on (the build skills, as prerequisites) — Python, API, prompting, RAG, agents.
- The AI systems you support (architectures) — single agent, multi-agent, agentic RAG, MCP, frameworks.
- Deploy & serve — FastAPI, Docker, env/secrets, probes, lifecycle, statelessness, CI/CD, change mgmt.
- Observe & monitor — tracing, the signals, logging, dashboards & SLIs, tooling.
- Reliability & incident response (SRE for AI) — resilience patterns + M31 (SLOs, on-call, incidents, runbooks, postmortems).
- Evaluate & assure quality — golden sets, scorers, eval gates in CI, regression detection.
- Secure & make safe — OWASP LLM Top 10, prompt injection, defenses, guardrails, M33 secret rotation.
- Optimize cost & performance — token cost model, caching, routing, trimming, streaming, batch.
- Data, memory & feedback loops — memory/state, M33 vector-store ops, the data flywheel, PII redaction.
- Support users & the business — agent UX/streaming, plus M32 (ticketing, support tiers, escalation, AIOps).
- Govern & continuously improve — responsible AI, governance frameworks, closing the loop.
- Capstones & practice — the build capstone, the complete-agent capstone, and the M31 ops-support drill.
Coverage at a glance
- 34 / 34 modules (M0–M33) are placed on the map, nothing is orphaned.
- 120 nodes across 13 branches.
- The original 25 gap proposals: 16 are now the built Part E modules (M31–M33); 9 remain as
[NEW]proposals to build out next.
See coverage.md for the node-by-node audit (auto-generated, so it can never drift).