## 🚀 Default Engagement Prompt

Copy and customize the template below to reliably activate AetherDeploy's full strategic and tactical depth.

---

**Subject: Production Deployment Architecture & 90-Day Roadmap for [Model / Use Case Name]**

**Model / Use Case**: [e.g., Fine-tuned Llama-3.1-70B RAG knowledge assistant for customer support, weekly-retrained LightGBM fraud model, or YOLOv8-based visual inspection system]
**Current Maturity**: [Validated offline on holdout / Packaged and running in staging / Currently serving X% of live traffic with observed issues]
**Business SLAs & Constraints**: [p99 latency < 220 ms at 800 QPS sustained, 99.95 % availability, monthly inference cost target <$4,800, SOC 2 Type II + GDPR + internal responsible-AI policy]
**Traffic Profile**: [Steady baseline with 3-4x daily peaks, or highly spiky marketing-driven loads, or batch nightly scoring of 12 M records]
**Infrastructure Reality & Preferences**: [Existing EKS cluster with GPU node pools and ArgoCD, or all-in on Vertex AI / SageMaker, or greenfield with strong open-source preference and limited platform headcount]
**Known Risks & History**: [Previous model experienced unexplained 4-hour degradation after a data schema change; prompt injection attempts logged in staging; finance team has flagged AI spend as top concern; regulatory audit scheduled in 11 weeks]
**Team & Process Context**: [3 MLEs + 2 platform engineers; GitHub Actions + ArgoCD; on-call rotation exists but limited experience with model-specific incidents]

**Requested Deliverable**:
Please produce a complete, board-ready production deployment architecture and phased 90-day implementation roadmap containing:

1. Target architecture (detailed component diagram in Mermaid or clear textual dataflow + rationale) together with explicit comparison against two credible alternative approaches in a trade-off matrix.
2. End-to-end CI/CD and promotion pipeline design including all automated test gates, security scans, canary analysis logic, and one-click rollback paths.
3. Production monitoring, drift detection, and automated response specification: exact metrics, tools, thresholds, dashboards, owners, and escalation runbooks.
4. Security, privacy, and responsible-AI controls specifically tailored to the data sensitivity and decision impact of this use case.
5. Phased rollout plan (shadow → canary → progressive → full) with quantitative entry and exit criteria for each phase and clear rollback triggers.
6. 12-month TCO model (compute, storage, networking, observability, human ops) plus three prioritized cost-optimization recommendations with projected savings and risk trade-offs.
7. Complete runbooks and documentation package that would allow a new SRE or MLE to understand, operate, and redeploy the system from scratch within one week.
8. Comprehensive risk register (likelihood × impact × mitigation owner × blast radius) and blast-radius analysis for the proposed design.
9. Executive summary (5-7 bullets) and suggested slide content suitable for presenting to VP Engineering, CISO, and business stakeholders.

Please begin by listing any critical clarifying questions or missing information that would materially change the recommendation. Once answered or assumed, deliver the full package in the order above.

---

**Pro Tips for Maximum Leverage**:
- Attach or paste any existing model card, evaluation report, current Kubernetes manifests, Terraform, cost dashboards, or recent incident timeline.
- State the regulatory classification explicitly (e.g., 'high-risk' under EU AI Act, 'consumer lending model' under OCC guidance).
- Be honest about timeline pressure and risk tolerance (e.g., 'we need something live in four weeks even if it requires more manual oversight initially; full automation can come in phase 2').

This prompt consistently unlocks AetherDeploy's complete depth in both strategic architecture and executable implementation detail.