ml-system-design
Use when designing end-to-end ML systems, choosing batch vs streaming inference, preventing training/serving skew, building data flywheels, or planning ML infrastructure scaling.
npx skills add majiayu000/claude-skill-registry --skill ml-system-design-jlaws-dotfiles --agent claude-code
Same command for any agent — swap --agent for codex, cursor, copilot.
Weekly change comes from our own snapshots, not the repository page — it measures attention, not adoption.
# ML System Design ## Serving Pattern Decision Table | Factor | Batch | Online (Real-time) | Streaming | |--------|-------|---------------------|-----------| | Latency tolerance | Hours | <100ms | Seconds | | Request pattern | Scheduled/bulk | Per-request | Continuous | | Freshness need | Stale OK | Must be fresh | Near-real-time | | Compute cost | Low (off-peak) | High (always-on) | Medium | | Example | Recommendation email | Search ranking | Fraud detection | | Infra complexity | Low | Medium | High | | Failure mode | Retry whole job | Per-request retry | Checkpoint + replay | ## System Design Template Work through these phases sequentially. Skip none. ### Phase 1: Problem and Metrics ``` 1. Business problem → ML problem mapping - "Increase engagement" → ranking/recommendation - "Reduce fraud" → binary classification - "Extract info" → NER/sequence labeling 2. Metrics - Business: revenue, CTR, churn rate - Model: precision, recall, AUC, NDCG - System: p50/p99 latency, throughput, availability 3. Constraints - Latency budget (e.g., <50ms for serving) - Cost ceiling (e.g., <$0.001 per inference) - Data privacy (PII handling, GDPR) ``` ### Phase 2: Data Pipeline ``` Raw Sources → In
- Serving Pattern Decision Table
- System Design Template
- Phase 1: Problem and Metrics
- Phase 2: Data Pipeline
- Phase 3: Feature Engineering
- Phase 4: Model Selection
- Phase 5: Training Pipeline
- Training/Serving Skew Prevention
- Data Flywheel Design
- Scaling Patterns
- Horizontal Serving
- Model Parallelism Decision
- Cost Modeling
- Monitoring
What does the ml-system-design skill do?
Use when designing end-to-end ML systems, choosing batch vs streaming inference, preventing training/serving skew, building data flywheels, or planning ML infrastructure scaling.
How do I install it?
Run `npx skills add majiayu000/claude-skill-registry --skill ml-system-design-jlaws-dotfiles --agent claude-code` — it drops the skill into your project so the agent can pick it up. Swap the --agent value for codex, cursor or copilot if you use one of those.
Where does this skill come from?
From majiayu000/claude-skill-registry, a repository with 534 stars. We read it straight from the repository tree rather than a submitted listing, so what you see here is what is actually published.
Is a popular skill a good skill?
Not necessarily. Stars measure attention, not adoption — a repository can trend for a week and be abandoned. That is why we show the weekly change from our own snapshots next to the total, instead of a single flattering number.