OPEN SOURCE

Building ML Infrastructure in the Open

Alliance AI contributes production-grade machine learning tools to the open-source community. Our frameworks power clinical research and enterprise ML systems.

100+
ML Models
15-Stage
AutoML Pipeline
MCP Native
Agent-Ready

Endgame

Production ML Framework

Production-aware machine learning under the scikit-learn API. Calibrated probabilities, interpretable models, deployment guardrails, and agent-ready via MCP.

PyPI version GitHub stars Downloads Python 3.10+ Apache 2.0 license
$ pip install endgame-ml[tabular]
Deployment Integrity

Calibrated probabilities, conformal prediction, leakage detection, latency constraints. Models built for production, not just accuracy.

Interpretable by Design

30+ glass-box models (EBM, GAM, CORELS, GOSDT, SLIM). Glass-box models match GBDT accuracy while remaining fully auditable.

Agent-Ready via MCP

Built-in Model Context Protocol server. AI agents (Claude, Copilot) can build, evaluate, and deploy ML pipelines through natural language.

Full AutoML Pipeline

15-stage pipeline with quality guardrails, HPO, ensembling, threshold optimization, calibration, and explainability. Time-budgeted.

Interactive Reports

Endgame Classification Report — 42 chart types showing model performance, calibration, and feature importance
42 chart types, self-contained HTML, no CDN dependencies View Demo Report →

PIE

Clinical ML Pipeline

Parkinson's Insight Engine — an end-to-end ML pipeline for the Michael J. Fox Foundation's PPMI dataset. Built on Endgame, PIE automates data loading, feature engineering, model training, and report generation for Parkinson's disease research.

MIT license MJFF Research Community

Pipeline Stages

Raw PPMI Data
Data Reduction
Feature Engineering
Feature Selection
Classification
Report
PPMI
Dataset Support
1 Command
Full Pipeline Run
6 Stages
With HTML Reports

Contribute

Both projects are open to contributions. Whether you're a researcher, ML engineer, or just curious — we'd love your input.