People are very good at learning new concepts after observing just a few examples. For instance, a child will confidently point out which animals are "dogs" after having seen only a couple of examples of dogs before in their lives.

The core of the Xyggy Engine is a novel machine learning algorithm grounded in the ability to learn concepts from examples and to generalize to new items - which is one of the cornerstones of intelligence.

Further, this learning and generalization happens in realtime.

From a query of two or more items, the Engine learns in realtime which other items belong to it, and return relevant results in ranked order.




As an example, a query with the two animated movies ‘Monsters Inc.’ and ‘Up’, would return other similar animated movies, like ‘Wall-E’.

With realtime-learning:

    1. Results are infered from all available data – which can be added, updated and deleted in realtime.

    2. Simplify the system architecture and save on recurring costs by eliminating batch-learning.

    3. Scale applications from one machine to thousands of cores on a distributed cluster.

Machine intelligence should be dynamic not static.

With the Xyggy API, items can be added to or removed from the query set at each invocation, to improve the relevance of results.

The API includes calls to find “more like these” or to act on "relevance feedback" which can be designed into the UI for an interactive and engaging UX.

Autonomous behaviour and engineered serendipity are easy to implement too.


Applications can range from a powerful “I am the query” personalization engine to recommendation systems.

From anomaly detection in machine data to fraud detection in transaction data.

Content-based search for images and audio ... and much more

For the first time, machines are making predictions and it signals the start of a momentous cultural shift. Driving this forward are machine learning algorithms with data as the fuel.

On a parallel path, the growing ubiquity of mobile devices is leading to a convergence between the physical analog world and the digital world.


The Xyggy purpose is very simple:

To make it easier and faster for developers to deploy a new class of machine intelligence ... with applications that are realtime, scalable and interactive on a secure fault-tolerant cloud


There is a lot more to the Engine's capability and these presentations will get you upto speed:

Start with the Engine overview followed by the Use-cases. A walk through the short Tutorial is highly encouraged and keep the API information handy as you do.

Anomaly Detection is a critical application area for machine learning, and images are used as the example to identify outliers.

To demonstrate massively scalable and realtime machine learning, Benchmarks were run on a large distributed system at the Cambridge University High Performance Computing Center.

Two key operations were timed - query execution and adding new items - as the number of data items and non-zeros increase across a rising number of cores from 16 to 1024.

Download the Free Trial Version and start to prototype machine intelligence applications.

Windows 32-bit xyggy-engine-trial-0.2.8-win32.tar.gz
Linux 32-bit xyggy-engine-trial-0.2.8-linux32.tar.gz
OSX 64-bit xyggy-engine-trial-0.2.8-osx64.tar.gz
  • Package installers will be made available in the future.
  • License and Readme in package.
  • If you have problems with a download then please contact us.


  • Along with the Get Started, review the Tutorial and the API information.

    If you already have feature engineered data then you'll be up and running in no time. Give it a go - it really is so much easier and faster. With the dynamic API, queries can be modified between invocations to deliver interactive applications.

    If you have questions or want to know more then please get in touch:

    Email Hello Xyggy, or at the Xyggy Google Group.