Personal roadmap and reference library for becoming an AI Systems Engineer for regulated European industries by September 2027. Covers 10 engineering disciplines, 3 flagship project blueprints, and a production-readiness methodology applied to the
andrelair-platformsovereign Kubernetes stack.
Live docs: https://andrelair-platform.github.io/aisystemmaster/
Platform docs: https://andrelair-platform.github.io/minicloud-platform-docs/
| Directory | Content |
|---|---|
AI_Systems_Engineer_Playbook.md |
Master career roadmap — 10 pillars, market demand, interview readiness, monetization path |
corpus-engineering.md |
Acquisition, OCR, chunking, deduplication, delta indexing |
model-selection.md |
12 model categories, decision framework, trade-off matrix |
llmops.md |
Prompt versioning, canary releases, inference servers, cost engineering |
ai-security-red-teaming.md |
OWASP LLM Top 10, attack taxonomy, Garak/PyRIT, mitigation by layer |
mcp.md |
stdio vs SSE transports, tool schemas, auth patterns |
cicd.md |
GitHub Actions AI pipeline, eval in CI, Docker, golden sets |
kubernetes-gitops.md |
ArgoCD App-of-Apps, Kustomize, GPU scheduling, k3s sovereign |
enterprise-security.md |
IAM, multi-tenant RBAC, Vault dynamic secrets, DLP, Cosign/SBOM |
reliability-engineering.md |
HA patterns, DR runbook, SLO/SLA, incident management |
compliance-governance.md |
DPIA, AI Act, data retention, audit log architecture |
project1-enterprise-knowledge-platform.md |
RAG at scale, eval CI, observability |
project2-workflow-automation-platform.md |
Agent orchestration, MCP, human-in-the-loop |
project3-insurance-compliance-copilot.md |
Sovereign stack, AI Act controls, audit lineage |
website/ |
Docusaurus 3.10.0 — published to GitHub Pages |
Every flagship project publishes these with full methodology:
- Groundedness rate (%)
- Citation accuracy (%)
- Refusal rate on out-of-scope (%)
- p50 / p95 / p99 latency (ms)
- Cost per resolved query (€)
- Hallucination rate on adversarial set (%)
Projects run on andrelair-platform — a 5-node bare-metal Kubernetes cluster (k3s + ArgoCD + Harbor + Vault + Authentik + Longhorn) simulating the IS of a French B2B insurer.
MIT — André Kanmegne