Why Jazz?

Efficiency is never wrong

“A lot of people think that performance is about doing the same thing, just doing it faster. And that's not true! If you can do something really fast, really well, people start using it differently."



Linus Torvalds, one of the few creators of game-changing tools, including Linux and git.

Navigating this site

This page is the short version on how we envision Jazz as a necessary platform to advance both AI research and deployment of AI in production. A longer, but still understandable without a technical background, explanation on why it is needed, can be found in the motive. How Jazz solves the identified opportunities for improvement over current platforms is discussed in how is Jazz a solution?

The problem we solve

Jazz is not just a step forward in performance; it aims at storing, processing and communicating data in such an efficient manner that it fundamentally transforms the way we use AI, by generating automated code to solve problems.

Past and present

Jazz has been under development since end of 2016 although a narrow version of the current design. It was released as OSS by BBVA in December 2017 and started its development towards new goals in the first quarter of 2018. You can also check the technical vision, the how to start and the programmer's documentation all of them updated continuously.

How we plan to make it happen

Jazz is built as open source software by a group collectively know as kaalam.ai. The group has already being delivering Jazz versions and documentation using automated integration since 2018. The initial declaration of intentions is here hello kaalam.ai!

And finally, the pitch!

  • We are building Jazz because we need Jazz and, more importantly, there is nothing like Jazz.
  • Jazz builds on top of a large OSS stack and can run the latest deep learning models, such as LLMs, as efficiently as it gets, since it includes onnx-runtime over an unparalleled memory mapped persistence. But that is not the important part or even the reason to build it. It can generate onnx on-the-fly, run it and evaluate its result, orders of magnitude faster than what any other system can.
  • Only with Jazz we can do the science, we want to do. During the last seven years we have done a lot of research on how to ground conceptual reasoning by making an algorithm generate code and evaluate its results. Jazz is the tool to bring these ideas from the lab into production. At some point in time, if you want to do particle physics you have to build an accelerator. In this analogy, Jazz is the accelerator.