Train your models in R, Scikit-learn or Apache Spark and use them for real-time scoring in enterprise Java applications.

usa 2017 tel-151813-3.jpg


Take your R, Scikit-learn, Apache Spark, TensorFlow and many other models from predictive analytics suites and convert them to standard PMML format.

usa 2017 tel-171153-Edit-3.jpg


Add our JPMML Evaluator library to your enterprise Java application, deploy a PMML model and get under 1ms evaluations per data point. Our PMML execution engine supports both batch and real-time scoring.

perjatsi lastega-5185-Edit-Edit-2.jpg

Evaluate in cloud

Use our REST API for PMML model deployment and evaluation that can be hosted in cloud and supports any client that speaks REST.

List of products

Evaluation Vertical

JPMML-Evaluator – Java Evaluator API for PMML. Supports PMML schema versions 3.0 through 4.3. Delivers millions of scores per second on Java/JVM platform.

JPMML-Evaluator for Apache Spark – Use JPMML-Evaluator as an Apache Spark
ML transformer.

JPMML-Evaluator for Apache Pig - Use JPMML-Evaluator as an Apache Pig UDF.

Openscoring REST service – REST over HTTP approach to make JPMML-Evaluator functionality available across all languages/platforms.

JPMML-Evaluator for Apache Hive - Use JPMML-Evaluator as an Apache Hive


Conversion Vertical

SkLearn2PMML – Convert your scikit-learn pipelines to PMML. Available as a Python package and standalone Java library.

JPMML-SparkML – Convert your Apache Spark pipelines to PMML. Available as
PySpark package and standalone Java library.

R2PMML – Convert your R models to PMML. Available as a R package and standalone Java library.

JPMML-TensorFlow - Converts TensorFlow linear and DNN regressors and classifiers to PMML.

JPMML-LightGBM - Convert LightGBM models to PMML. Available as Python and R plugin, and standalone Java library.

JPMML-XGBoost - Convert XGBoost models to PMML. Available as Python and R plugin, and standalone Java library. 

Here's the deal:


Openscoring/JPMML software is dual-licensed under AGPLv3 and commercial license (usually BSD 3-Clause License).

If you're create and license your software under AGPLv3 - you're more than welcome to use our software free of charge.

If you're in the business of creating commercial software, we can offer our products under more relaxed licensing conditions, most commonly under BSD 3-Clause License, for a fee.

Here's the quick difference of AGPL and BSD licenses:

Users of AGPLv3-licensed software: Users of BSD-licensed software:
MUST disclose source Don't have to disclose source
CANNOT sublicense Can sublicense
AGPLv3 conditions APPLY to software as a service (if the server is AGPLv3, the client MUST be too) Don't have constraints related to SaaS

There're different opinions on the Internet about what AGPLv3 license allows and what it doesn't*, but let us be honest with you - we expect you to buy a license if you're using our software for commercial purposes :-)

Commercial licenses are available in two terms:

Annual (you pay a yearly fee)

Perpetual (you pay one-time fee)

Your license is per 'covered work' - in simple terms a software project with all the dependencies. We try to keep it simple and don't charge per users, nodes, servers, etc.

The annual/perpetual license fee is per covered work.

Our software always comes with all updates and upgrades included during the licensing period (you're right - if you buy a perpetual license, you're covered forever).

Contact for quote!

* In our view - if your software doesn't function properly without ours, it's functional dependence and thus the clients must be AGPLv3 without re-licensing agreement.


Support & Services


We take pride in writing software that virtually doesn't need any support. But if you still discover a rare issue then quick and timely support is provided on the:

JPMML mailing list

We're also working on commercial support offer - feel free to contact us ( and let's discuss your requirements.

We've also helped our clients to integrate JPMML tools to their machine learning pipelines. Again, feel free to drop us an email if you need a bit of consulting.


Latest Articles on Our Blog

A Few Words About Openscoring


Openscoring was founded in 2014 when Villu Ruusmann discovered the lack of Java based tools for evaluating PMML models during writing his PHD thesis. Soon that void was filled and Openscoring was born.

We're a smallish (but growing!) Estonian company. That doesn't stop us from regularly dealing with customers included in NASDAQ100.

We always go an extra mile if we can be more transparent and open.


Key People

Villu Ruusmann

Founder & CTO

Villu is the one who is likely to answer your technical questions.

Karel Kravik

Partner & COO

Karel is the one who will pick up the business leads and quote requests.

PS: You're absolutely free to use the direct email addresses should you hate the anonymous info@ just as we do!