4 Minute Demo of the Tecton Feature Platform

David:
Hey, and welcome to this walkthrough of Tecton. Tecton is a feature platform for machine learning. It’s built to make it easy to get the data that you need to your machine learning models. Tecton really helps accelerate two workflows, that we’re going to cover in this video. The first is feature production, where Tecton is going to help your team build the data pipelines that compute features and keep them up to date. And the second is feature consumption, where Tecton provides a set of tools to discover features as well as simple APIs to collect features for training models or doing model inference. Let’s get into it.
Tecton provides a simple and powerful set of tools to build the data pipelines that power ML applications. With Tecton, feature pipelines are declared in a simple declarative code framework. Tecton then creates and orchestrates those data pipelines in your data platform of choice, something like Snowflake or Databricks.
Importantly, Tecton’s data pipelines can also operate on realtime data from your application, meaning that you can build features once and use them both for training and serving machine learning models.
Tecton’s feature framework is built in Python. It is meant to be used by both data scientists and engineers. The declarative code first approach to authoring features allows for easy integration into most applications. Features can be managed in source control, and rolled out to production with CICD.
Once you’ve built some feature pipelines, Tecton makes it simple to consume those features for ML. To start, Tecton provides you with best in class capabilities to discover features and share them with the rest of your team. You shouldn’t need to reinvent the wheel every time you want to build a new feature.
Once you’ve picked the features you need to train a model, our Python SDK makes it simple to generate training data. With one API call, Tecton can join together your features, and our time travel capabilities ensure you avoid common pitfalls like data leakage.
After you’ve trained and deployed an ML model, Tecton provides a highly performance serving layer, optimized to deliver features to production systems. Simply invoke Tecton’s REST API to retrieve the features you need for real-time machine learning in a handful of milliseconds.
That’s all for this quick demo. To summarize, Tecton is built to support the end-to-end machine learning data lifecycle. If you’d like to try it out, you can find out more on our website.