Everything We Do is Motivated by Real-World Experience

Author: Donald Thompson
Everything We Do is Motivated by Real-World Experience

We build the Maana Knowledge-Centric Technology Platform by tackling the biggest problems of the world’s largest companies, so everything we do is motivated by real-world experience.  We are proud to announce our first major version increase of the platform that incorporates technologies and lessons learned over the course of nearly 4 years.  But instead of just rattling off a list of new features and improvements, I thought I’d share more about why we did what we did and present a more technical perspective.

We have found that to solve real projects requires providing capabilities to three groups of people: technical experts, subject-matter experts, and knowledge workers. And, if you will recall, we refer to Maana as knowledge-centric technology, which means we focus on Knowledge models (both data and computational) as our core abstraction and the rest of the technology we build is for the creation, management, adaptation, utilization, reuse, etc. of these models.

For technical experts, such as data analysts, information modelers, data scientists, software engineers, Maana provides a wealth of low-level capabilities from interactive REPLs, shell scripts, interactive “workbooks,” reusable libraries, REST APIs, extensibility points, etc.  But, in this release, we’ve introduced a major extensibility mechanism and higher-value functionality with what we call Knowledge Assistants.  The goal with Knowledge Assistants is to strike the balance between people and machines, which we call user-guided, machine-assisted, in using artificial intelligence, machine learning, and many other techniques to augment users as they extract, explore, discover, and model new forms of knowledge describing some aspect of the business domain.  These Knowledge Assistants support activities such as finding connections between different data silos; finding complex similarities between entities like jobs, projects, vendors, customers, patients, etc.; mining unstructured documents, like failure reports, contracts, operating manuals, for data tables, key concepts, relationships, and facts; utilizing time-series search and analysis.  This list of assistants goes on and will quickly become a rich area for advanced algorithms and experiences to turn raw data into usable knowledge.

While technical in nature, many of the Knowledge Assistants are often very useful to subject-matter experts, better enabling them to self-service (or become what some call citizen data scientists).  But we are also introducing a separate set of experiences we call Knowledge Applications that are more tailored for the business user, covering areas like predictive maintenance, supply chain optimization, accounts receivable optimization, etc.  These experiences can be thought of like solution templates with data models, computational models, and a dedicated UX, that can be connected to the relevant sources of knowledge that have been defined by the solution (e.g., invoices, sensor data, maintenance logs).  These experiences can easily be extended to suit any custom requirements.

The biggest technical update comes to our core Knowledge Graph, which now has its own high-speed distributed storage and query engine for graphical and time-series data with support for multiple indexes.  The Knowledge Graph sits at the center of any Maana solution and must provide interactive results for a variety of demanding use cases, in addition to providing policy-enforced access controls, multi-datacenter federation, and other features that made it hugely challenging to continue building on layers of 3rd party components.  In addition to streamlining support for all of our storage operations, the performance numbers see orders of magnitude improvement from the previous version with all common operators implemented in native code running close to the data, minimizing round-trips and key exchanges between nodes.

Moving outward from the Knowledge Graph, this release of Maana has a completely revamped the distributed compute orchestration layer which allows Maana to expand beyond Spark as the only distributed compute backend to being able to support any backend, such as Google’s TensorFlow, gridMathematica, or even a proprietary system through a pluggable interface.  This allows Maana to support using the right tool for the job, whether it is a Python-based optimization package or a complex deep learning network, Maana now supports fluid integration and composition.

Continuing the discussion on integration, we’ve embraced the fact that most companies have already made investments in various infrastructure, tools, and applications and that any real-world system must be a good citizen in this ecosystem.  As such, Maana continues to invest in support for the Hadoop family of technologies, but also support for ODBC, OData, REST/JSON, and other standard integration protocols enabling support for 3rd party ETL, data wrangling, visual analytics, dashboards, etc.

All of this is experienced within a revamped user experience that fully supports knowledge modeling, extremely powerful semantic search and exploration, rich, automatic data visualizations, and fast access to Knowledge Assistants.

Please stop by our Booth #452 at Strata+Hadoop World in NYC this week for a demo and deeper discussion.  Here is additional information on the Maana Knowledge Platform.



Palo Alto, CA
Houston, TX
Bellevue, WA
London, UK
Dhahran, KSA

Strategic Partners

  • Accenture
  • Microsoft Azure

Learn More

Connect with Us

Stay in the know with the latest information about Maana services, events, news and best practices by email.