AIDRIN Documentation
AIDRIN (AI Data Readiness Infrastructure) is an open-source tool for evaluating whether a dataset is ready for artificial intelligence and machine learning work. It scores quality, structure, fairness, and privacy risk, and returns the results as plots and machine-readable JSON. Use it from a browser, a terminal, a notebook, or Claude Code.
It reads CSV, Excel, JSON, NumPy (.npz), HDF5, and Parquet, plus Croissant, RO-Crate, DCAT-US 3.0 and 1.1, and
DataCite JSON for metadata.
Six Dimensions of Data Readiness
AIDRIN groups its metrics into six dimensions. Each is documented in Web Application Usage; Metric Names Across Interfaces maps every metric between the web interface, the command line, and the Python library.
Dimension |
What it measures |
|---|---|
Data Quality |
Completeness, duplicates, and outliers, including row-level and temporal completeness and custom validity rules |
Impact of Data on AI |
Feature correlation and how strongly each feature relates to your target |
Fairness and Bias |
Class imbalance, representation rates, and statistical parity across sensitive attributes |
Data Governance |
Re-identification risk: k-anonymity, l-diversity, t-closeness, entropy risk, HIPAA identifiers, and differential privacy |
Understandability and Usability |
Variable-unit metadata readiness and FAIR compliance against the Croissant, RO-Crate, DCAT-US 3.0 and 1.1, and DataCite schemas |
Data Structure |
Constant features, collinearity, skewness, and kurtosis |
|
|
Four Ways to Use AIDRIN
- Web Interface
An interactive, browser-based dashboard. Upload a dataset, select dimensions and metrics, and explore results with visualizations and downloadable reports — no coding required. Available hosted at aidrin.org or self-hosted locally. See Web Application Installation and Web Application Usage.
- Command Line Interface (CLI)
Run data readiness metrics directly from your terminal. Suitable for automated pipelines, CI workflows, and headless environments. Also includes an agentic evaluation component for domain-aware data readiness question answering and remediation grounded in scientific literature. See CLI Installation and CLI Usage.
- Python Library
Import AIDRIN into notebooks and scripts. A per-metric functional API for exploratory work, and a registry-driven headless API that shares its metric set with the command line for pipelines. See Library Usage.
- Claude Code (MCP)
Ask Claude Code to assess your dataset in plain language. AIDRIN ships an MCP server (
aidrin-mcp) and a Claude Code skill that together let Claude run metrics, interpret results, and write a readiness report — with no commands to remember. See AIDRIN Skill.
Metrics
Web Interface
Python Library
Developers