A public development program from engineering leader to Head of AI / CAIO (product/tech track), aimed at enterprise and regulated / government environments. Learning in the open (build in public): a knowledge base plus capstones with measurable quality and governance.
Principle: not the path of an execution engineer, but leader-grade depth — enough depth to make architectural decisions, hire and evaluate engineers, and answer to the board and the regulator.
| # | Area | Level | Progress |
|---|---|---|---|
| 01 | LLM Architectures | BUILD | ✅ |
| 02 | Context Engineering | BUILD | ⬜ |
| 03 | RAG Frameworks | BUILD | ⬜ |
| 04 | Agentic Architectures | BUILD | ⬜ |
| 05 | Service Management | LEAD | ⬜ |
| 06 | Safety & Reliability | BUILD ×2 | ⬜ |
| 07 | Observability & Testing | BUILD ×2 | ⬜ |
| 08 | Scalable Production | LEAD | ⬜ |
| 09 | Development Essentials | LEAD | ⬜ |
| 10 | Optimized Deployment | LEAD | ⬜ |
| 11 | Multimodal Mastery | selective | ⬜ |
| 12 | Ecosystem Adaptation | habit | ⬜ |
| 13 | Career / Build in Public | in parallel | ⬜ |
Progress: ⬜ not started · 🟧 in progress · ✅ closed (the area's checklist is complete).
Governance, Risk & Compliance — EU AI Act, NIST AI RMF, OWASP LLM Top 10, model risk, data residency. The main differentiator for a regulated / government role.
| # | Capstone | Areas | Launch | Status |
|---|---|---|---|---|
| 01 | Enterprise RAG assistant | 01–03, 06, 07, 09 | make up && make seed && make test (requires key+Postgres) |
code + full code review in docs/CODE_REVIEW.md |
| 02 | Multi-agent workflow | 04, 05, 06, 07, 08 | make eval · make test (offline, no key) |
eval gate PASS: 90% block-rate, 100% tasks |
| 03 | Self-hosted / VPC deployment | 10, governance | make check · make decide (offline) |
residency-checker PASS/FAIL + TCO break-even |
Quick start without a key or DB (capstones 02 and 03 run offline):
cd capstones/02-multiagent-guardrails && make eval # attacks + tasks + latency gate
cd capstones/03-self-hosted-deployment && make check # data-residency verdict
make test-all # tests for all capstonesCI (GitHub Actions) runs tests across all three capstones and the offline eval on every push.
knowledge-base/ 13 areas: theory, examples, exercises
governance/ risk/compliance layer + a small in-house framework
capstones/ flagship projects with evals and a governance passport
- Go through the areas in tier order (BUILD 01–04, 06, 07 first).
- For each area, complete its readiness checklist in the README.
- Put exercise results in
examples/, and the numbers in the README. - Take the area's quiz (
quiz/QUIZ.md, scored 0–100) — a score ≥ 80 closes the area (✅ in the table above). See self-assessment. - Finish each capstone with a governance passport and a retrospective (area 13).
Regional requirements (e.g. language) are omitted — the focus is on transferable competencies.
License: MIT. Sensitive/confidential data is not committed to the repository (see .gitignore).