Skip to main content

​
Overview

Generate a comprehensive MLOps Strategy document that defines how ML/AI models will be developed, deployed, monitored, and governed throughout their lifecycle. Ensures AI systems are reliable, reproducible, and compliant with governance requirements.

​
When to Use

Use /arckit.mlops when your project includes:
  • Machine Learning models (classification, regression, NLP, computer vision)
  • Large Language Models (LLMs) or Generative AI
  • Algorithmic decision-making systems
  • AI-assisted automation
Run after:
  1. /arckit.requirements - understand ML use cases
  2. /arckit.data-model - understand training data
  3. /arckit.ai-playbook - governance context (UK Gov)

​
Usage

/arckit.mlops SageMaker for fraud detection model
/arckit.mlops Vertex AI GenAI chatbot with LLM fine-tuning
/arckit.mlops Azure ML classification pipeline

​
What It Creates

File: projects/{project}/ARC-{PID}-MLOP-v1.0.md Sections:
  1. ML Use Cases - Model purpose, inputs, outputs, performance targets
  2. Model Lifecycle - Training, validation, deployment, monitoring, retirement
  3. MLOps Maturity - Level 0-4 assessment and target state
  4. Training Pipeline - Data prep, feature engineering, training, evaluation
  5. Serving Infrastructure - Batch vs real-time, scaling, latency SLAs
  6. Feature Store - Feature engineering, storage, serving
  7. Model Monitoring - Drift detection, performance degradation, fairness
  8. Model Registry - Versioning, lineage, metadata
  9. Responsible AI - Bias testing, explainability, human oversight
  10. Governance - UK AI Playbook, ATRS, MOD JSP 936 (if applicable)
  11. Cost Management - Training costs, inference costs, optimization

​
MLOps Maturity Levels

LevelCharacteristicsAutomation
Level 0Manual process, scripts, notebooksNone
Level 1ML pipeline automationTraining
Level 2CI/CD for ML pipelinesTraining + Deployment
Level 3Automated monitoring and retrainingFull lifecycle
Level 4Self-healing, auto-scaling, A/B testingAutonomous

​
Supported ML Platforms

  • AWS SageMaker - Managed training, hosting, Pipelines
  • Google Vertex AI - Unified ML platform, AutoML, Workbench
  • Azure Machine Learning - Designer, Pipelines, Endpoints
  • MLflow - Open source experiment tracking and serving
  • Kubeflow - Kubernetes-native ML workflows

​
UK Government AI Compliance

Integrates with:
  • /arckit.ai-playbook - 10 AI principles, 6 ethical themes
  • /arckit.atrs - Algorithmic Transparency Recording Standard
  • /arckit.jsp-936 - MOD AI assurance (defence projects)

AI Playbook

UK Government AI Playbook compliance

DevOps

CI/CD pipelines for model deployment

FinOps

ML infrastructure cost optimization

Data Model

Training data schema and governance

Build docs developers (and LLMs) love

Get started for freeTalk to us