AI Design Patterns: Patterns & Tactics for AI-Enabled Systems

Generic, brand-neutral engineering capability; deep, sourced playbooks live in references/ and knowledge/. [DOC]

Generic, brand-neutral engineering capability; sourced playbooks in references//knowledge/. [DOC]

TL;DR

AI design patterns define reusable solutions to recurring architectural problems in AI systems. This skill produces a pattern selection analysis covering maintainability tactics, availability tactics, AI-specific patterns (Feature Store, Champion-Challenger, Shadow Deployment, Drift Detection), traditional patterns adapted for AI, anti-pattern detection, and a decision framework that maps system requirements to recommended patterns [EXPLICIT]

When to Use

• Selecting design patterns for new AI system architecture
• Evaluating existing AI system against known patterns and anti-patterns
• Designing fault detection and recovery for AI systems (drift, model failure, data quality)
• Implementing feature store, champion-challenger, or shadow deployment
• Defining maintainability and availability tactics for AI pipelines
• Planning migration from ad-hoc ML code to pattern-based architecture
• Reviewing anti-patterns and proposing remediation strategies

When NOT to Use

• Internal module structure and layer architecture -> ai-software-architecture
• CONOPS and operational concept -> ai-conops
• Pipeline design and CI/CD -> ai-pipeline-architecture
• Testing strategy -> ai-testing-strategy
• GenAI/LLM-specific patterns (RAG, agents) -> genai-architecture
• Traditional software patterns without AI context -> software-architecture

---

Sub-capabilities (resource map)

Deep, evidence-tagged playbooks — open the one the task needs (ICM Layer 3, on-demand). [INFERENCE]

Procedure

1. Resolve the sub-capability; open the matching references/ playbook. [EXPLICIT]
2. Apply its decision tables; pick the strategy explicitly. [EXPLICIT]
3. Validate against the Quality Criteria and tag every claim. [EXPLICIT]

Quality Criteria

• Sub-capability resolved to one playbook. [INFERENCE]
• Claims evidence-tagged. [EXPLICIT]

Contract

• Aceptación: capability resolved to its reference playbook, applied, validated, evidence-tagged. [EXPLICIT]
• Límites: · Focuses on patterns and tactics, not pipeline design (see ai-pipeline-architecture) · Does not design internal module structure (see ai-software-architecture) · Does. [EXPLICIT]
• Casos borde: Early-Stage System with One Model: Most patterns are overkill for a single-model MVP. Start with Drift Detection and basic monitoring. Add Feature Store only when a second mode. [EXPLICIT]
• Supuestos: · System has or will build AI pipeline infrastructure (not running ad-hoc scripts) · Team understands the distinction between model development and model operations · Infrastructur. [SUPUESTO]
• Trade-off: Pattern Enables Constrains When to Use --- --- --- --- Feature Store Consistency, reuse, drift monitoring Infrastructure overhead, governance cost Multiple. [EXPLICIT]

Packet

Capas del packet, cargables bajo demanda (disciplina ICM: una capa por vez, nunca todas juntas): references/ guías de profundidad (cargar UNA por etapa) · knowledge/ cuerpo de conocimiento · prompts/ prompts listos · examples/ salida de ejemplo · agents/ subagentes del packet · assets/ recursos estáticos.