Sisense Hunch – Leadership Through Radical Innovation l Sisense

Today we are proud to announce the launch of one of the most radical innovations introduced in the Big Data market in years – Sisense Hunch – a Data Cognition Engine.

Sisense Hunch is a new category of an ultrahigh performant database and analytics engine based on deep neural networks. Hunch delivers extremely high query rate performance (microsecond to milliseconds) over very large datasets compressed by orders of magnitude (terabytes down to megabytes) with a pre-defined accuracy loss.

We have been developing Hunch under the radar over the past couple of years and at first we just couldn’t believe our own eyes – the results were just too good to be true!

When we tested Hunch on real datasets the results simply wowed us. For example, we have built a 2MB Hunch model (what we like to call a Hunch “brain”) comprising only 500 neurons replacing one billion records and 250GB. The resulting Hunch brain delivers results at a rate of 100,000 answers per second and with over 99% accuracy.

How can Sisense Hunch be used?

Well, there are tons of possibilities that Sisense Hunch opens up. Here are a few:

Make Big Data Interactive

Users can enjoy interactive data exploration over terabytes of data using Sisense Hunch’s microsecond to millisecond response times, only querying the more expensive, slower Big Data system directly when very precise detail is required.

Bring the Value of Big Data to the Edge

Sisense Hunch compresses terabytes of data into a Hunch brain that occupies 2MB-5MB for each terabyte. Once in place, Hunch requires no access to the underlying data, putting the power of your data anywhere you need it while eliminating the need for storage, processing power, and bandwidth.

Guaranteed 100% Data Privacy

Since Sisense Hunch doesn’t retain any knowledge about the lowest level of detail in your data. There is zero risk that queries could return sensitive data that is prohibited by your policy or regulatory compliance requirements.

Zero Footprint Big Data “Storage”

Go ahead and drain your data swamp. In cases where the row-level detail is not needed or cannot be stored (such as with GDPR), Hunch can completely replace Big Data repositories with a nearly accurate solution that will satisfy most analytic needs.

Innovation Design Partner – Celestica

Headquartered in Toronto, Celestica is one of the world’s largest electronics manufacturing services (EMS) company. Like Sisense, Celestica is committed to delivering the highest level of service to their clients at the most competitive price. By testing the ability to deploy Sisense Hunch directly into their manufacturing process, Celestica was able to demonstrate it can deliver significant improvement to the speed and level of accuracy for quality testing on their production line.

One of many services that Celestica provides is the manufacturing of electronic circuit boards that are built into products that we all use every day. The equipment which manufactures these circuit boards is highly specialized and is programmed to detect the rare errors that inevitably occur when manufacturing something that requires such a high degree of precision.

These built-in error detection algorithms work very well. However, Celestica is known for their commitment to operational excellence and they understand that every small improvement means better outcomes for their clients both in quality and in cost. Over the years Celestica has amassed huge amounts of data about these machines and boards. However, querying for the status of each board and comparing it to the vast trove of historical data in order to detect potential issues with it as it is being manufactured isn’t realistic given the vast amount of data and the timescale needed for an answer versus query time.

Sisense Hunch was able to convert this vast amount of data, containing over 1.9 billion records, into a 2.5MB neural network enabling edge analytics at the production machine level in real-time. Sisense Hunch provides the equivalent of a full database query in 0.1 milliseconds with better than 99% accuracy compared to the original dataset for every board as it is being manufactured. This one problem resolution is worth hundreds of thousands of dollars for Celestica for this use case alone. Other use cases exist and are now being considered for future implementations.

We have a hunch that you’re ready to jump in…

Sisense Hunch is now at a stage where we are opening it up to our customers, seeking additional use cases, and looking to work with new design partners. If you have read until here and your mind is racing with the possibilities that Hunch opens up to you, reach out to us!

We imagine the impossible and deliver the improbable. Sisense Hunch is only the beginning!