Teradata launches Elastic Python Compute on AWS & Azure

Teradata has launched Elastic Python Compute within its AI Studio platform. The product is now generally available on AWS and Microsoft Azure.

The tool gives data science teams a managed Python environment for running artificial intelligence and machine learning workloads across distributed systems. It is designed to let users process large datasets, train models and run custom Python applications without setting up and managing their own Ray clusters.

Elastic Python Compute is built on Ray, an open-source distributed computing framework, and is integrated with Teradata's Autonomous Knowledge Platform and AI Studio. Workloads run within the Teradata platform rather than requiring data to be moved to external environments.

This approach is aimed at organisations that want to keep data and model development within existing governance and residency controls. The system supports local model hosting and avoids external API dependencies and token-based inference costs.

Data handling

The platform reads and writes directly to the Teradata Database, S3 object storage and open table format catalogues, including Apache Iceberg and Delta Lake. It also supports files and images for model training and scoring.

Users can deploy the tool on a single server or across clusters, with separate clusters available for different workloads. Teams can choose CPU or GPU resources depending on the task across cloud, on-premises and hybrid environments.

Teradata also highlighted support for Arrow Flight SQL, which allows parallel columnar data transfer between applications and the Teradata platform. Benchmark testing showed speeds of up to 84 times faster than row-based transfer methods, although results may vary by environment.

Customer examples

Teradata cited a retail bank that had been exporting tens of millions of customer records into local Python environments for credit feature engineering. Those workflows now run inside the Teradata environment, with credit scoring pipelines running three times faster and providing a full audit trail from raw data to final model output.

In telecommunications, the company described a provider whose data science team had been exporting billions of call detail records and network telemetry events to external Python clusters. After moving the work into Teradata's platform, feature engineering runtime fell from four hours to 20 minutes, while model accuracy improved by 12 per cent because the provider could train on the full dataset rather than a sample.

Teradata also pointed to a government maritime agency using the tool to combine satellite imagery, radar signals, AIS data and commercial imaging to reconstruct vessel movements. The analytics helped improve visibility into ships attempting to conceal their routes.

Executive view

Teradata positioned the launch as a response to a longstanding issue for enterprise data science teams, which often export data from core platforms into separate Python environments for model building. That process can create governance, sovereignty and operational risks, particularly in regulated sectors.

"Enterprise data science teams have spent years working around the fundamental constraint of moving data outside their platforms to build and train AI models, accepting the governance risk and cost that comes with it. Elastic Python Compute removes that constraint. Teams can now run distributed AI workloads at scale where their data already lives, and get to production faster without compromising on governance or cost control," said Sumeet Arora, Chief Product Officer at Teradata.

Google Cloud availability is expected later, while an on-premises version is already available through Teradata Factory.