On August 10, Microsoft announced the release of ‘SynapseML’, which is an open-source library that helps simplify creation of machine learning pipelines.

SynapseML is a machine learning library built on Apache Spark which makes it easier to train production-ready models that classify and regress anomaly detection, translation, speech to text, and various other ML challenges. The new platform has created an automated SparkML binding generation system that translates SparkML and SynapseML APIs into other programming languages like Python, R, and .NET.

SynapseML integrates various ML technologies such as LightGBM, Vowpal Wabbit, ONNX, and the Cognitive Services into one API compatible with MLFlow. The library is available in different languages like Python, Scala, Java, and R. 

The simplification of different unified ML learning frameworks with a single, scalable API that is data- and language-agnostic will allow users to either stream or batch serve applications and focus on high-end structuring of their data and tasks.

SynapseML v0.10.0 is usable from .NET, C#, F#. Here is an example of training a distributed LightGBM model: 

Additionally, SynapseML is now expanding its 50 Cognitive Services with Azure Open AI Service that will tap into 175-Billion parameter language models (GPT-3) from OpenAI that can create text and generate code near human parity.

SynapseML will now fully support ‘MLflow’, a platform for managing machine learning lifecycle which allows users to track experiments, pack code into reproducible runs and deploy models. 
Along with this, the inclusion of Binder technology inclusion will allow users to experiment with SynapseML without setup, install, infrastructure, or Azure account.