NEW: "IntegratedML Custom Models" Early Access Program -- Deploy Your Python ML Models in SQL

We're excited to announce the Early Access Program for IntegratedML Custom Models, a powerful new capability coming in IRIS 2026.1!

What Is It?

IntegratedML Custom Models extends the existing IntegratedML/AutoML feature by letting you deploy your own custom Python ML models directly within SQL queries. While IntegratedML AutoML (and H2O and DataRobot providers) test a select set of models against the given dataset and then chooses one of them, Custom Models gives you full control—custom preprocessing, any scikit-learn compatible model or Python class that implements the scikit-learn standard methods, and third-party libraries like Prophet or LightGBM—all executing in-database without data movement.

Why It Matters

• Full Control: Write custom preprocessing, feature engineering, and model training logic
• Same SQL Interface: Use familiar CREATE MODEL, TRAIN MODEL, and PREDICT() commands
• In-Database Execution: No data export/import cycles, real-time predictions on live data
• Works with AutoML: Use the right tool for each use case—AutoML for quick models, Custom Models for advanced scenarios

Quick Example

CREATE MODEL CreditRiskModel
PREDICTING (default_risk)
FROM CreditApplications
USING '{"model_name": "CustomCreditRiskClassifier", ...}'

TRAIN MODEL CreditRiskModel

SELECT customer_id, PREDICT(CreditRiskModel) as risk_score
FROM NewApplications

Who Should Join?

This is an open Early Access Program. Join if you:

• Have ML/data science experience with IRIS
• Want to deploy custom Python models in production
• Can dedicate ~10-15 hours over 8-12 weeks
• Are excited to shape the future of IntegratedML

What's Included

✅ 4 complete demo applications (Credit Risk, Fraud Detection, Sales Forecasting, DNA Similarity) ✅ Comprehensive documentation (installation, troubleshooting, API reference) ✅ Direct feedback channel to the product team ✅ Early access before IRIS 2026.1 GA release

Requirements

• IRIS 2025.2+ with IntegratedML installed
• Python 3.8+ (3.11+ recommended)
• macOS, Linux, or Windows
• Target: <30 minute installation

How to Join

1. Sign up: Visit https://evaluation.intersystems.com and register for the IntegratedML Custom Models EAP
2. Review the repository: https://github.com/intersystems-community/integratedml-custom-models
3. Get started: Follow the installation guide and start exploring

What to Expect

• Email interaction: The InterSystems Product team will engage with you via email throughout the program
• User survey: We'll ask you to complete at least one survey at the end of the EAP to share your experience and feedback
• Your feedback matters: Your input directly shapes the final product before GA release

Timeline

• EAP Launch: November 2025
• EAP Duration: 8-12 weeks
• Target GA: IRIS 2026.1 (Q2 2026)

Questions?

• Sign up: https://evaluation.intersystems.com
• Repository: https://github.com/intersystems-community/integratedml-custom-models
• FAQ: https://github.com/intersystems-community/integratedml-custom-models/blob/main/docs/EAP_FAQ.md

Your feedback will directly shape this feature before it launches to all IRIS users!