Tecton

Tecton Feature Platform FAQ

Basics

Tecton is a feature platform. Our feature store is focused on storing data offline and online, keeping data consistent, and providing an API to consume the data. We also allow teams to define and automate data transformation pipelines that compute and materialize batch, streaming, and real-time features. Additionally, the platform includes built-in monitoring.

Tecton is designed for machine learning engineers, data engineers, and data scientists looking to use a feature platform to build optimized batch, streaming, and real-time data pipelines for ML-driven use cases.

The primary use cases for Tecton are real-time ML-driven applications, including recommendation systems, fraud detection, dynamic pricing, search ranking, marketing personalization, or real-time loan approvals. The feature platform can be used for nearly any use case that requires machine learning to drive real-time product decisions.

Technical

Feature logic in Tecton is code-based as opposed to being tuned within the UI. Users can define feature logic in SQL, PySpark, SnowPark, or Python in a notebook or any other Python environment, and manage and version control the associated files in a version control repository (e.g., git) of your choice.

Yes, Tecton provides on-demand transformations. Alongside the feature store, feature repository, and feature monitoring, on-demand transformations are a key component of the Tecton feature platform. They’re commonly used when feature transformations cannot be pre-computed but must be computed at inference time.

Tecton has native support for time-window aggregations that are most commonly seen in machine learning use cases (min, max, count, sum, variance, etc.). Tecton’s implementation allows the end user to easily trade off between feature freshness, cost-efficiency and read latency. Batch, streaming, and real-time features are all supported. If Tecton’s native aggregations don’t suffice, you can always fall back to registering arbitrary SQL transformations as batch features or to computing features outside of Tecton and ingesting them into Tecton for serving.

Yes, Tecton’s SDK helps mitigate feature leakage. It accomplishes this by managing feature and data timestamps, and data delay modeling.

Deployment

Tecton is available on AWS and GCP. Azure support will be available soon. Please get in touch with us directly if you want access to an early version.

Not currently. As a SaaS solution, Tecton can run in your cloud environment, but not in an air-gapped environment. The reason for this is that supporting availability SLAs requires that we have access to manage it.

You don’t need to set up an ETL pipeline to start using Tecton. Tecton automates that for you.

The default is to store offline data in the user’s account and online data in Tecton’s account. We offer other deployment models where data can be stored entirely in the user’s account.

Tecton supports Redis and DynamoDB on AWS and Redis and BigTable on GCP. The right choice depends on your use case. We typically help teams think it through based on the scale and data profile of their use cases.

In Tecton, different git repos can map to different Tecton workspaces that can operate in parallel. You can pull from multiple workspaces in production. Ideally though, we encourage all data dependencies for any software, like an ML model, to be defined within a single repo to avoid dependency problems.

Security

Tecton offers the option of authenticating users via SSO. Access can be granted to users, teams, or services using Access Control Lists (ACLs). SSO and ACL can be used by our customers to better manage GDPR compliance requirements. Tecton as an organization is SOC 2 Type 2 compliant as a key aspect of holding a high bar to protect our customers’ environment and data.

Data is encrypted at rest and in public transit using industry-standard AES-256 encryption.

The Tecton risk team performs an annual risk assessment to identify risks related to the industry and those specific to Tecton. The risk areas in the assessment are ranked by likelihood, impact, and risk level (high, medium, or low). Formal risk-treatment plans are created for any identified risks and are tracked throughout the year to continuously drive down risk at the organization.

Still have questions about Tecton?

Get in touch with us!

Request a demo

tecton-logo-white

Get Your Features to Production

Book a Demo

Unfortunately, Tecton does not currently support these clouds. We’ll make sure to let you know when this changes!

However, we are currently looking to interview members of the machine learning community to learn more about current trends.

If you’d like to participate, please book a 30-min slot with us here and we’ll send you a $50 amazon gift card in appreciation for your time after the interview.

CTA link

or

CTA button

Contact Sales

Interested in trying Tecton? Leave us your information below and we’ll be in touch.​

Unfortunately, Tecton does not currently support these clouds. We’ll make sure to let you know when this changes!

However, we are currently looking to interview members of the machine learning community to learn more about current trends.

If you’d like to participate, please book a 30-min slot with us here and we’ll send you a $50 amazon gift card in appreciation for your time after the interview.

CTA link

or

CTA button

Request a free trial

Interested in trying Tecton? Leave us your information below and we’ll be in touch.​

Unfortunately, Tecton does not currently support these clouds. We’ll make sure to let you know when this changes!

However, we are currently looking to interview members of the machine learning community to learn more about current trends.

If you’d like to participate, please book a 30-min slot with us here and we’ll send you a $50 amazon gift card in appreciation for your time after the interview.

CTA link

or

CTA button