Backprop is a serverless model platform that makes it simple for developers to use machine learning models in any application.
Using models is done via
Tasks, such as
Question answering and
Tasks act as a middleman between a request and a model,
which makes it easy to use a variety of models without having to worry about the specifics of each.
As part of Backprop, we include a curated and evaluated list of pre-trained models that can solve a variety of text and vision based tasks. These are the latest and greatest models from the top AI research teams. See an overview of them in the Model Hub.
With the right tasks and models, it is easy to build scalable intelligent solutions like conversational chatbots, semantic text/image search, email classifiers, and more.
Every model hosted on Backprop behaves like an isolated serverless function. This means that whenever you perform a request, it is not slowed down by any other traffic.
This allows for keeping track of usage purely on a computation time basis, without having to put arbitrary prices on tokens and individual models.
With this approach, anyone can upload their own model that will function as seamlessly as the pre-trained models. Learn more about it in the uploading tutorial.