Announcing Tecton on Google Cloud: Accelerate the Development of ML-Powered Applications

Announcing Tecton on Google Cloud:

Accelerate the development of ML-powered applications with Tecton on Google Cloud

Tecton on Google Cloud provides an easy and fast way to build, process, manage, and share high-quality ML features across your organization to power your ML applications. The Tecton Google Cloud partnership allows customers to accelerate the time to value for ML models, maximize model performance and reliability, and control costs for their ML applications.

Data is the lifeblood of any machine learning application. To make more accurate predictions, ML models need access to high-quality data signals, commonly referred to as ML features. These ML features are usually characterized by refined data with high signal value that incorporate all the latest information available to an organization.

Predictive ML applications are widely deployed in the enterprise and driving business value in the real world today. These applications are often powered by hundreds or thousands of ML features. Use cases include detecting fraud at time of transaction, recommending the next best purchase for a customer, or providing a customized insurance quote to a new prospect at sign-up. In all these instances, the higher the quality of the ML features, the higher the business value generated.

The data challenge in production ML today 

Today, for many organizations, the time to value for machine learning is often far too long as it commonly takes months to deploy an ML model to production. Once it is in production, model predictions are frequently incorrect or experience downtime and missed SLAs. These challenges are exacerbated as organizations look to scale the number of ML models in production. ML features are at the heart of the problem.

ML teams struggle to build and manage production ML feature data pipelines. These pipelines need to process and transform raw data from batch, streaming, and real time sources into ML features, ensure these features are readily available for accurate training and serving, and monitor their performance in production. They also need to run cost efficiently on flexible infrastructure while meeting stringent performance and scale requirements. And they need to be managed across models in an auditable, shareable, and scalable way.

How Tecton on Google Cloud helps ML teams

Google Cloud provides both advanced data infrastructure and industry-leading services for building and running ML-powered applications. Tecton accelerates the building, processing, sharing, and serving of ML features across your ML applications on top of Google Cloud.

Tecton is a feature platform that automates the steps involved to build and manage production grade ML features. Here is how it works:

• Data teams can easily define features as code using Tecton’s declarative framework
• Under the hood, Tecton orchestrates the physical data pipeline required to transform and materialize ML features
• Features are stored in a low-latency online store for real-time serving, and an offline store for training dataset generation and offline inference
• Once in production, Tecton monitors feature data quality and operational service levels

Tecton is designed to meet enterprise scale and security requirements, supporting hundreds of thousands of queries per second with low latencies. Tecton is SOC 2 Type 2 compliant and enables GDPR compliance. With Tecton’s hybrid SaaS deployment, customers can keep control over their data while benefiting from the agility of a SaaS solution.

With Tecton on Google Cloud, customers can easily iterate on feature development, build state of the art models, and get them to production quickly and reliably. Specifically, Tecton allows you to:

Accelerate the time to value

Automatically deploy production machine learning data pipelines and quickly iterate on your ML models

Maximize performance and reliability

Increase your model accuracy and serve features at low latency and massive scale with enterprise-grade SLAs

Control costs

Avoid overhead of building and maintaining your own infrastructure and control spend

Future-proof your ML stack

Be prepared for the shift toward real-time predictions and generative AI

Accelerate the development of ML applications with Tecton

Predictive ML is widely used to power many use cases in production today, including fraud, recommendations, and personalization. With Generative AI and LLM applications, ML features are used to complement a user input in the LLM prompt, providing the context for LLMs to generate more accurate and personalized outputs.

Tecton on Google Cloud is a great option for building production-grade ML features. As a single hub for ML features, it enables data teams to easily power both their predictive and generative ML applications.

We thank the Google Cloud team members who co-authored the blog: Banruo Yu, Technical Account Manager, Google Cloud; Christian Williams, Principal Architect, Google Cloud; Raj Goodrich, Cloud Customer Engineer, Google Cloud; Sridhar Kavikondala, Cloud Customer Engineer, Google Cloud