Help us improve
Share bugs, ideas, or general feedback.
From systems-design
Designs end-to-end ML system architectures for problems like recommendations, fraud detection, search ranking, and content moderation. Outputs diagrams, data flows, tech stacks, trade-offs, and phased plans.
npx claudepluginhub melodic-software/claude-code-plugins --plugin systems-designHow this skill is triggered — by the user, by Claude, or both
Slash command
/systems-design:ml-pipelineThis skill is limited to the following tools:
The summary Claude sees in its skill listing — used to decide when to auto-load this skill
Design an end-to-end ML system architecture for a given problem.
Designs production ML systems from data ingestion and feature stores to model training, serving, and monitoring. Use for ML pipelines, MLOps infrastructure, and system design interviews.
Turns model work into production ML systems with data contracts, repeatable training, quality gates, deployable artifacts, and monitoring. Useful for ranking, search, recommendations, classifiers, forecasting, embeddings, LLMs, anomaly detection, and batch analytics.
Builds production ML systems with PyTorch 2.x, TensorFlow 2.x, JAX, Hugging Face, and more. Covers model serving, feature engineering, A/B testing, monitoring, and ML infrastructure.
Share bugs, ideas, or general feedback.
Design an end-to-end ML system architecture for a given problem.
$ARGUMENTS - The ML problem to design for (e.g., "recommendation system", "fraud detection", "search ranking", "content moderation")
Clarify requirements by understanding:
Load relevant skills based on the problem:
ml-system-designllm-serving-patternsrag-architectureml-inference-optimizationvector-databasesSpawn the ml-systems-designer agent for comprehensive design:
Design the complete pipeline:
Address cross-cutting concerns:
/sd:ml-pipeline recommendation system for 100M users
/sd:ml-pipeline real-time fraud detection for payments
/sd:ml-pipeline search ranking for e-commerce with 10M products
/sd:ml-pipeline content moderation for social media
/sd:ml-pipeline ad click prediction at 1M QPS
/sd:ml-pipeline customer churn prediction
/sd:ml-pipeline demand forecasting for inventory
| Category | Key Considerations |
|---|---|
| Recommendations | Cold start, real-time signals, A/B testing |
| Fraud/Risk | Low latency (<100ms), rules + ML hybrid, feedback loops |
| Search/Ranking | Multi-stage ranking, personalization, position bias |
| NLP/LLM | Inference cost, caching, streaming responses |
| Computer Vision | GPU inference, batching, edge deployment |
| Time Series | Feature freshness, windowing, seasonal patterns |
A comprehensive ML system architecture including: