Skip to content

FeatCopilot

🚀 Next-Generation LLM-Powered Auto Feature Engineering Framework

Automatically generate, select, and explain predictive features using semantic understanding of your data

Benchmark Highlights

  • +197% Max Improvement

    Simple models benchmark on delays_zurich dataset

  • +420% with LLM Engine

    LLM-enhanced feature generation boosts results

  • 48% Datasets Improved

    Tabular engine improves 20/42 datasets tested

  • +8.55% AutoML Best

    FLAML benchmark improvement on abalone dataset

View Full Benchmark Results

Two Modes of Operation

Sub-second feature engineering using rule-based transformations:

from featcopilot import AutoFeatureEngineer

# Fast, deterministic feature engineering
engineer = AutoFeatureEngineer(
    engines=['tabular'],
    max_features=50
)
X_transformed = engineer.fit_transform(X, y)  # <1 second

Best for: Production pipelines, real-time inference, reproducible results

Domain-aware semantic feature generation with any LLM provider:

from featcopilot import AutoFeatureEngineer

# LLM-powered semantic features
engineer = AutoFeatureEngineer(
    engines=['tabular', 'llm'],
    max_features=50
)
X_transformed = engineer.fit_transform(
    X, y,
    column_descriptions={'age': 'Patient age in years'},
    task_description='Predict heart disease risk'
)  # 30-60 seconds

Best for: Exploratory analysis, domain-specific features, maximum accuracy

What is FeatCopilot?

FeatCopilot is a Python library for automated feature engineering powered by large language models. It analyzes column meanings and descriptions to generate domain-aware features, applies intelligent selection to keep only the most predictive ones, and provides human-readable explanations for every feature it creates.

  • Multi-Engine Architecture

    Tabular, time series, relational, and text feature engines in one unified API

  • LLM-Powered Intelligence

    Semantic feature discovery, domain-aware generation, and automatic code synthesis

  • Intelligent Selection

    Statistical testing, importance ranking, and redundancy elimination

  • Sklearn Compatible

    Drop-in replacement for scikit-learn transformers in your ML pipelines

Why FeatCopilot?

Feature FeatCopilot Featuretools TSFresh AutoFeat OpenFE CAAFE
Tabular Features ✅ ❌ ❌ ✅ ✅ ✅
Time Series ✅ ❌ ✅ ❌ ❌ ❌
Relational ✅ ✅ ❌ ❌ ❌ ❌
LLM-Powered ✅ ❌ ❌ ❌ ❌ ⚠️
Semantic Understanding ✅ ❌ ❌ ❌ ❌ ⚠️
Code Generation ✅ ❌ ❌ ❌ ❌ ⚠️
Sklearn Compatible ✅ ✅ ✅ ✅ ✅ ❌
Interpretable ✅ ⚠️ ⚠️ ⚠️ ❌ ✅

Installation

# Basic installation
pip install featcopilot

# With LLM capabilities
pip install featcopilot[llm]

# Full installation with all extras
pip install featcopilot[full]

Getting Started

License

FeatCopilot is released under the MIT License.