Skip to content

Featrix SDK: Deep Dive

This page is for the folks who want to know more about what's inside. You don't have to remember any of this to be successful with Featrix.

Powered by Vector Embeddings

Featrix is powered by vector embeddings, which enables Featrix to tackle the hardest AI problems in a robust manner. This provides a huge leg up on other techniques for prediction, classification, and clustering of data.

Today vector embeddings are a bit mysterious, which is why 80+ people recently came to hear our CTO discuss what are embeddings and how do they work. You can check out this blog post to learn more.

featrix-tech-stack
compare-tickets

For all developers

Most AI SDKs are a lot of work. You'll need some stats books to go along with the documentation. Tuning hyperparameters, cleaning data, adjusting data, overtraining and underfitting--these are all issues you'll tackle head on with the other options on the market.

But not with Featrix. We've built our SDK to enable any developer to build powerful AI models and run them with an authenticated API endpoint.

You wouldn't set up your own cloud, write your own database, or write your own compiler for commercial applications these days. The most productive organizations use these tools off the shelf--and the same should be true for AI.

Built with the best

Our system is built on PyTorch and leverages the best tooling and techniques available. We have taken a data-centric approach to building our AI system so you can skip the data prep and get right to work.

Lightning Fast

Our inference latency is fantastic. We do all the caching and optimization for you so you just have to worry about the application you're building. Leave the AI to us.

Flexible Deployments

Our API is open: you can review the code, port it to new platforms, and more. The backend can be licensed with source code and you can run Featrix models outside of our environment with an enterprise license.

 

featrix-sphere-small

98% less code than doing things by hand with other solutions

The below code is a 15 line solution to the "spaceship Titanic" Kaggle competition. Try it out for yourself.

kaggle-98-percent
One solution to this problem included over 10 pages of data wrangling to achieve the same results we get with Featrix.

Ready to get started?

How about a poster?

AI is a confusing space, so we built a glossary poster.

You can download the PDF or we'll be happy to mail you a 22" x 28" copy for your wall.

Get the poster today!

poster-top