Senior ML Engineer
Production ML engineering skill covering model deployment, MLOps pipelines, LLM integration, RAG systems, and model monitoring. Provides step-by-step workflows, code templates, and tooling scripts to operationalize models.
✨ What it does
- Model deployment workflow with canary releases and validation thresholds
- MLOps pipeline setup with feature stores, experiment tracking, and retraining triggers
- LLM integration with provider abstraction, retry logic, and cost management
- RAG system implementation with vector DB selection and chunking strategies
- Model monitoring with drift detection and alert thresholds
- Scripts for deployment, RAG scaffolding, and monitoring setup
🎯 When to use it
- Deploying trained ML models to production with monitoring
- Setting up MLOps infrastructure like MLflow, Kubeflow, or Kubernetes
- Monitoring model performance or detecting data drift
- Building RAG pipelines or integrating LLM APIs with retry and cost controls
🚀 How to use
Trigger by asking about MLOps, model deployment, monitoring, RAG, or LLM integration. Provide your specific scenario (e.g., framework, scale, constraints). The skill offers workflows, code templates, and scripts. Example prompts:
Help me deploy my PyTorch model to production with monitoring.
Set up an MLOps pipeline with MLflow and a feature store.
How do I integrate an LLM API with retry logic and cost controls?
📄 Output: Deployment artifacts (Dockerfile, Kubernetes manifests), MLOps pipeline configurations, RAG system scaffolds, monitoring dashboards, and code templates.
📦 Add this skill to Claude Code
# 1. Get the skills repo
git clone --depth 1 https://github.com/alirezarezvani/claude-skills /tmp/claude-skills
# 2. Copy this skill into your project (or ~/.claude/skills for all projects)
mkdir -p .claude/skills
cp -r /tmp/claude-skills/engineering-team/skills/senior-ml-engineer .claude/skills/senior-ml-engineerSkill source: engineering-team/skills/senior-ml-engineer/SKILL.md
⚠️ Good to know
Focuses on production and operational concerns, not model research or initial training; assumes existing trained models.
❓ FAQ
What validation thresholds are recommended for model deployment?
p95 latency should be under 100ms and error rate under 0.1% before promoting to full production.
How should I handle LLM cost management?
Do not hardcode prices; look up current rates at request time and use tier ratios (small, mid, large) for model routing decisions.
What triggers automated retraining?
Scheduled cron jobs, performance drops below accuracy threshold, data drift with PSI > 0.2, or new data volume thresholds.
🤖 Overview, features, install steps and FAQ were generated from the project's SKILL.md on Sep 4, 2026. Always check the original source before running commands.