How do AI agents orchestrate, control, and stay reliable?
AI agents and orchestration that move from prompt to outcome. Each system in this domain ships with a Mermaid architecture diagram, a numbered implementation map, and a checkmark list of documented build phases. The original source document is kept per system.
| Reader | What to look at | Time |
|---|---|---|
| Recruiter | This README, badges | 10 sec |
| Hiring Manager | Systems list below, per-system Validation | 2 min |
| Engineer | Per-system Architecture + Implementation | 15 min |
| Learner | Per-system end-to-end source content | 30 min |
AI and ML engineering across agents, multi-agent orchestration, model integration, and self-applied daily-driver tooling. Includes forges, evaluation harnesses, and the meta-tooling that produces other systems.
What it isn't. A guarantee of agent reliability under all conditions. A vendor-neutral comparison of AI frameworks.
- Build an Agent Control Plane: Fail-closed AI boundary proven by an airgap test: 11 refusals break to 0, then restore.
- Build an AI Job Application Pipeline: Dual-model council where Codex flagged all 12 claims before any package shipped.
- Architect Your Manifesto: 607 atomic claims gated by confidence, with Gemini counsel kept model-family independent.
- Ship a Repo Explainer Pipeline: Nine agents on typed JSON contracts with council veto over a deterministic render stack.
- Personal AI Research Agency: Capability routing with adversarial verification that caught a live prompt injection.
+ 5 other systems in the full catalog: INDEX.md.