The Cognitive Framework; it's all in the (source) code

In this short post I want to discuss an important deployment feature from Chatterbox Labs which has generated great partner feedback, and that really differentiates Chatterbox Labs' Cognitive Framework: we deploy our source code. This codebase represents over 6 years of academic research and development within an enterprise grade software framework.

Unlike a Software as a Service (SaaS) delivery model, where you pay to access software that is running in the cloud, accessed either in the browser or via a set of APIs, Chatterbox Labs deploy our technology on premise at a partner site, with complete source code access. This is not a huge deployment process; it takes a maximum of 2 days to complete. Once this is done, our partners have complete access and modification rights to our codebase.

Our partners find this really important for a number of reasons, three of which being:

  1. Data Privacy
    Our Cognitive Framework is used to build machine learning, text classification products in days, often based upon highly confidential client data. Due to strict confidentiality requirements, there is no way that this data could be sent to a cloud service. As Chatterbox Labs technology is deployed inside the firewall, the data never leaves the partner's secure environment and with source code access the partner can carry out a full security audit should they wish to.

  2. Customization
    One size doesn't fit all. Our Cognitive Framework offers all the building blocks needed in order to build new machine learning classifiers. It follows a coding paradigm which has extensibility at its heart; if the partner wishes to build on top of our code base to add their own bespoke functionality, that's fine and baked into the Python software design of the framework. There’s no need to take on the whole codebase, the software is designed to ensure that you only need to modify the bits you care about.

  3. Latency
    Chatterbox's Java prediction engine is designed to operate in a real-time environment, making use of massive parallelisation across multicore server environments. Benchmarked on standard commodity hardware, the software is able to process 148k messages per second using 9 Java threads.
    With the round trip time inherently present in sending data over the public web, these types of speeds can only be dreamed of by using web APIs without investing millions of dollars into infrastructure and maintenance.

These are just the three of the benefits that I often hear from our partners when discussing source code deployment. If you would like to know more about our Cognitive Framework contact Andrew Watson, VP Strategic Alliances (andrew@chatterbox.co).