skills.rest
kalilurrahman
kalilurrahmanCOMMUNITY
Explore all skills by this author →
Website𝕏@krahmanGitHub

ml-pipeline

Design end-to-end machine learning pipelines covering training, deployment, and drift monitoring.

1|Updated Mar 21, 2026
Search Tags:#monitoring#fastapi#ml pipeline#machine learning#model deployment#mlflow#data drift
One-click install
npx skills add https://github.com/kalilurrahman/kr-claudiator-skills-original-prompts --skill ml-pipeline-kalilurrahman
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: ml-pipeline
Source: https://github.com/kalilurrahman/kr-claudiator-skills-original-prompts/tree/main/04-ai-ml/ml-pipeline
Command: npx skills add https://github.com/kalilurrahman/kr-claudiator-skills-original-prompts --skill ml-pipeline-kalilurrahman

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve? Building a production ML system involves far more than training a model — teams struggle with experiment tracking, reproducibility, deployment, and detecting performance degradation in production. This Skill provides a complete blueprint for designing an ML pipeline from raw data to monitored production deployment. ## Core Features & Use Cases - Full Pipeline Design: Covers data extraction, feature engineering, time-based train/val/test splits, training with MLflow experiment tracking, evaluation, deployment, and monitoring. - Production Deployment Patterns: Includes model registry promotion, FastAPI serving with health checks, Docker packaging, and Kubernetes deployment guidance. - Monitoring & Retraining: Implements drift detection with Evidently, Prometheus metrics, and Airflow-triggered automated retraining on performance degradation. - Use Case: A data science team building a churn prediction system can follow this Skill to set up XGBoost training with MLflow tracking, deploy a real-time prediction API, and automatically retrain when AUC drops more than 10% from baseline. ## Quick Start Ask the AI to design an end-to-end ML pipeline for your use case, specifying the problem type, data sources, and deployment requirements.

Frequently Asked Questions about ml-pipeline

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I build an end-to-end machine learning pipeline?▼

Structure the pipeline in five stages: data extraction and feature engineering, training with experiment tracking via MLflow, multi-metric evaluation, deployment through a model registry with FastAPI serving, and production monitoring with drift detection and automated retraining.

How do I track ML experiments with MLflow?▼

Wrap training in mlflow.start_run(), then log parameters, metrics, figures, and the model artifact. Register the best model in the MLflow Model Registry and promote it to the Production stage for serving.

How do I deploy a machine learning model as an API?▼

Load the production model with mlflow.pyfunc.load_model, expose a /predict endpoint with FastAPI and Pydantic request validation, add a /health endpoint, and package the service in a Docker container for Kubernetes deployment.

How do I detect data drift in production ML models?▼

Use Evidently's DataDriftPreset and TargetDriftPreset reports comparing reference training data against current production data. Alert when more than five features show drift, and trigger retraining when AUC degrades over 10% from baseline.

Why use time-based train test splits for ML?▼

Time-based splits prevent data leakage in temporal datasets by ensuring training data precedes validation and test data. Random splits would leak future information into training, producing overly optimistic evaluation metrics.

What should a machine learning model card include?▼

A model card documents model type and version, intended use and out-of-scope applications, training data sources, performance metrics, known limitations, and ethical considerations such as bias audits and excluded sensitive attributes.

Related Skills

Explore More Data & Analytics Skills→
Adelie-SquadAdelie-Squad
17

data-engineer

Clean and transform raw datasets into analysis-ready tables with pandas pipelines.

Official
Advanced
cgRGMcgRGM
1

churn-prevention

Design churn prevention systems with cancel flows, save offers, and dunning steps.

Community
Advanced
glinckerglincker
35

devops-specialist

Automates DevOps tasks including multi-cloud deployments, IaC, container orchestration, and monitoring setup.

Official
Advanced
duylinhdang1998duylinhdang1998

ml-infrastructure

Build and deploy ML training pipelines with Docker, Kubernetes, and FastAPI.

Community
Advanced