home / skills / pluginagentmarketplace / custom-plugin-python / machine-learning

machine-learning skill

safe

This skill helps you build and evaluate machine learning models in Python using scikit-learn, PyTorch, and TensorFlow end-to-end.

npx playbooks add skill pluginagentmarketplace/custom-plugin-python --skill machine-learning

Review the files below or copy the command above to add this skill to your agents.

Files (6)

SKILL.md

1.3 KB

---
name: Machine Learning
description: Python machine learning with scikit-learn, PyTorch, and TensorFlow
version: "2.1.0"
sasmp_version: "1.3.0"
bonded_agent: 03-data-science
bond_type: PRIMARY_BOND

# Skill Configuration
retry_strategy: exponential_backoff
observability:
  logging: true
  metrics: model_accuracy
---

# Python Machine Learning Skill

## Overview
Build machine learning models using Python libraries including scikit-learn, PyTorch, and supporting tools.

## Topics Covered

### Scikit-learn
- Data preprocessing
- Model selection
- Training pipelines
- Cross-validation
- Hyperparameter tuning

### PyTorch Basics
- Tensor operations
- Neural network modules
- Training loops
- DataLoader usage
- GPU acceleration

### Feature Engineering
- Feature selection
- Dimensionality reduction
- Feature scaling
- Encoding techniques
- Missing data handling

### Model Evaluation
- Metrics selection
- Confusion matrix
- ROC curves
- Learning curves
- Model comparison

### MLOps Basics
- Model serialization
- Experiment tracking (MLflow)
- Model versioning
- Serving models
- Reproducibility

## Prerequisites
- Python fundamentals
- NumPy and Pandas
- Statistics basics

## Learning Outcomes
- Train ML models
- Evaluate model performance
- Build ML pipelines
- Deploy models to production

Overview

This skill teaches practical Python machine learning with scikit-learn, PyTorch, and TensorFlow, plus supporting tools for production workflows. It focuses on end-to-end model development: data preparation, modeling, evaluation, and basic MLOps. Content is hands-on and aimed at developers who know Python and core data libraries.

How this skill works

The skill walks through preprocessing, feature engineering, and model pipelines using scikit-learn for classical ML tasks. It covers PyTorch fundamentals for neural networks—tensors, modules, training loops, DataLoader patterns, and GPU usage—and shows integration points with TensorFlow where relevant. It also introduces evaluation techniques, hyperparameter tuning, and simple MLOps: serialization, experiment tracking, and serving.

When to use it

Building classical ML models for structured data with scikit-learn
Developing and training neural networks with PyTorch for custom architectures
Preparing data pipelines: scaling, encoding, imputation, and feature selection
Comparing models using cross-validation, ROC/Learning curves, and metrics selection
Deploying, versioning, or tracking experiments for reproducible results

Best practices

Start with clear data preprocessing: handle missing values, scale features, and encode categoricals before modeling
Use pipelines to encapsulate preprocessing and model steps to avoid leakage and simplify deployment
Prefer cross-validation and stratified splits for robust performance estimates on imbalanced data
Profile training on CPU vs GPU and use DataLoader with proper batching and augmentation for efficient PyTorch training
Persist models and artifacts with consistent versioning and log hyperparameters and metrics for reproducibility

Example use cases

Predicting customer churn with a scikit-learn pipeline and hyperparameter tuning
Training an image classifier using PyTorch with custom Dataset and DataLoader
Reducing dimensionality with PCA or feature selection for faster model training
Tracking experiments with MLflow and exporting a model for inference in a web service
Comparing logistic regression, random forest, and a small neural network on the same preprocessing pipeline

FAQ

What prerequisites do I need?

Familiarity with Python, NumPy, Pandas, and basic statistics is recommended.

Will I learn deployment?

Yes. The skill covers basic model serialization, experiment tracking, and patterns for serving models, suitable for simple production workflows.