Teradata deepens Dataiku integration to accelerate enterprise AI projects

Today, multicloud data giant Teradata announced it is expanding its integration with AI startup Dataiku to enable enterprise users to import and operationalize their Dataiku-trained AI models within the Teradata Vantage platform.

The move, Teradata claims, will help companies move past deployment complexities and accelerate their AI projects from pilot to production, at scale. This is crucial as AI projects often end up in the proof-of-concept graveyard, and if they do make it to deployment, they are delayed by months due to operationalizing roadblocks. 

The expanded integration capabilities are available starting today, both companies said.

Teradata-Dataiku team up for AI

With Vantage, Teradata offers enterprises a modern analytics platform that combines open source and commercial analytic technologies to operationalize insights from data and enable descriptive, predictive and prescriptive analytics. Dataiku, on the other hand, gives a central working environment to experiment with data and train, deploy and manage AI applications.

While each platform has its own domain, they share synergies through the Teradata plugin for Dataiku. The integration allows their joint customers to access and execute certain analytic functions that reside in Teradata Vantage within Dataiku. This way, users of Dataiku could easily tie Vantage analytic functions, like data preparation, into their data science and AI project workflows. 

Now, Teradata and Dataiku are deepening this engagement by expanding the support for all Vantage analytic functions, including data cleansing, feature engineering, machine learning (ML), time series and digital signal processing. More importantly, the integration also now supports high-performance processing of Dataiku-developed ML models within Teradata Vantage. 

Previously, the models from the platform had to be converted to a common interchange format such as PMML (predictive model markup language). Now, the models can be imported in Dataiku’s own native model format, reducing the steps in the ML prediction pipeline and removing potential model conversion complexities. This can help teams accelerate their AI projects to production. 

“In general, data scientists will perform preparation, cleansing and transformation of their Vantage data through Dataiku workflows using the Teradata plugins … Analytic ML models are then trained with Dataiku ML algorithms using this training data from Vantage,” Hillary Ashton, chief product officer at Teradata, explained while speaking with VentureBeat. “These models can then be exported to Vantage in native Dataiku format for at-scale inference/scoring using ClearScape Analytics’ BYOM (bring your own model) functionality. This process can be iterated until a final model is achieved, with the Dataiku trained model productionized using BYOM model scoring in Vantage workflows,”

Available right away

Ashton said the enhanced capabilities are now live and multiple joint customers are already using them. She did not share specific outcomes seen so far, but said the company would be “happy to share results once this work is complete.”

Teradata’s Q1 2023 recurring revenue grew 4% in constant currency, which contributed to generating more than $300 million in gross profit and over $100 million in free cash flow. In December 2022, Gartner named the company a leader in its Magic Quadrant for cloud database management systems, citing price predictability and financial governance as key strengths.