Splice Machine introduces ML Manager Beta program to meet growing demand for operational AI

Splice Machine launched on Monday its beta program for ML Manager, a native data science and machine learning platform. Operating on top of Splice Machine's data platform, ML Manager empowers data science teams to maximize the performance of their machine learning models by removing the latency associated with building complex data pipelines, performing cumbersome transformations and training models on updated data.

With ML Manager, data scientists can experiment on ten times as many data pipelines to optimize their models, and when they eventually deploy the model into the production environment, it can immediately start making predictions in real time.

ML Manager provides data scientists and data infrastructure administrators with data science libraries, including MLlib, and familiar tools, such as Apache Zeppelin, Scala, Python, SQL and R. In addition, ML Manager delivers superior performance through seamless integration of Splice Machine tables and Spark DataFrames.

To simplify the machine learning lifecycle process, ML Manager integrates with MLflow (beta), an open source platform for managing the end-to-end machine learning lifecycle. With MLflow, data scientists can determine the effectiveness of their models by logging the parameters, code, and metrics for each experiment. Data scientists can then visualize and compare the various experiments, and ultimately deploy the model with the highest predictive power into production.

MLflow also provides data science managers with model governance and audit capabilities through visibility into datasets, transformations, and parameters that underlie the model deployed into the production environment.

With these enterprise-grade capabilities in place, MLflow-packaged models can then be deployed via Amazon Sagemaker for seamless implementation. MLflow and Sagemaker, as part of ML Manager, makes it possible for data scientists to build a higher number of effective models by spending less time on the challenges of experimentation, tracking, and deployment.  

IT data infrastructure consisting of separate transactional, analytical, and data science platforms is not viable for today's modern machine learning-powered applications because it is not agile enough to deliver the real-time data that is needed to continuously train the machine learning model in minutes versus hours or days.

ML Manager helps enhance the productivity of data scientists by running the machine learning models at the database level using the most current data available. This approach facilitates retraining the model whenever the underlying data attributes change, ensuring that the model is operating at its optimum performance.

"At Euler, we see first-hand how enterprises have tried to integrate machine learning and artificial intelligence into their business. Often times, companies are held back by their existing data infrastructure and the inherent latency it adds into the process," said Himanshu Nautiyal, CEO at Euler Systems. "With ML Manager as part of Splice Machine's operational AI data platform, companies now have a modern option to power applications and take control of machine learning on a unified platform."

"Enterprises can no longer afford to have latency in the machine learning models they rely on to power their mission-critical applications," said Monte Zweben, CEO, Splice Machine. "Whether you are relying on real-time data to combat online fraud, or want to serve up relevant recommendations to customers in real time, the need for reliable, timely data is omnipresent across industries. ML Manager removes the friction and delays in the data science process, allowing for optimal performance through fast, continuous learning and training of ML models."