How does the skill handle model versioning and rollbacks?

It uses a structured registration system that tracks performance metadata and training history, allowing for safe, automated rollbacks to previous versions if production metrics decline.

Does it help with capacity planning?

Yes, the load testing feature benchmarks your model's latency and error rates across different traffic volumes to recommend the maximum RPS per instance.

Which ML frameworks does this skill support?

It supports a wide range of popular frameworks including scikit-learn, XGBoost, and PyTorch, generating optimized deployment code tailored to each library's requirements.

Can I use this for both real-time and batch processing?

Yes, the skill provides specific patterns for REST APIs (real-time), batch predictors (scheduled via Airflow or Kubernetes), and streaming interfaces (Kafka/Kinesis).

What metrics are tracked in the generated monitoring setup?

It configures tracking for prediction latency (P50/P95/P99), throughput, error rates, prediction distribution, and feature drift detection.

ML Deployment Helper

Name: ML Deployment Helper
Author: anton-abyzov

byanton-abyzov

•

データサイエンスとML

Automates the transition of machine learning models into production-ready services with APIs, containerization, and monitoring.

概要

ML Deployment Helper is a specialized MLOps skill for Claude Code that bridges the gap between model development and production. It automatically generates production-grade deployment artifacts including FastAPI REST endpoints, batch predictors, and streaming interfaces. The skill ensures models are production-ready by handling containerization via Docker, setting up monitoring dashboards with Prometheus and Grafana, implementing A/B testing logic, and performing load testing to establish performance benchmarks, ensuring every model follows industry-standard MLOps best practices.

主な機能

Containerization with optimized Dockerfiles for various ML frameworks and Python versions
A/B testing infrastructure with traffic splitting and statistical significance tracking
Comprehensive monitoring setup including Prometheus metrics and Grafana dashboards
Performance benchmarking with automated load testing and RPS recommendations
Automated REST API generation using FastAPI with Pydantic input validation
13 GitHub stars

ユースケース

Establishing MLOps monitoring to detect feature drift and performance degradation
Converting a trained scikit-learn or XGBoost model into a scalable microservice
Setting up an A/B test to compare a new model version against a production baseline

概要

主な機能

Containerization with optimized Dockerfiles for various ML frameworks and Python versions
A/B testing infrastructure with traffic splitting and statistical significance tracking
Comprehensive monitoring setup including Prometheus metrics and Grafana dashboards
Performance benchmarking with automated load testing and RPS recommendations
Automated REST API generation using FastAPI with Pydantic input validation
13 GitHub stars

ユースケース

Establishing MLOps monitoring to detect feature drift and performance degradation
Converting a trained scikit-learn or XGBoost model into a scalable microservice
Setting up an A/B test to compare a new model version against a production baseline