Cisco is today open-sourcing the MindMeld platform for data scientists to build a complete end-to-end conversational AI assistants, bots, and applications for enterprise customers. MindMeld has been used by brands like Starbucks and Uniqlo to make conversational agents as well as by InQTel, the CIA’s venture capital arm.
The platform made its debut in November 2016, and just months later, MindMeld was acquired by Cisco for $125 million. MindMeld was used to create Cisco’s Webex assistant, which made its debut last month.
MindMeld’s Python-based machine learning framework comes with documentation, can create bots that support rich media, and come with blueprints, which act as scaffolding for data scientists who want to get started quickly or tackle particular use cases. Food ordering, collaboration, home assistant, and media discovery templates at launch based on customer feedback.
Former MindMeld CEO and Cisco cognitive collaboration CTO Tim Tuttle told VentureBeat in a phone interview said the platform will be available for free on Cisco’s DevNet portal and includes a user manual with instructions to get started.
“Most NLP platforms really only solve half the problem. They allow you to build machine learning models that can understand entities and do simple dialog management, but for most conversational applications that’s not enough. You can’t do NLP and dialog management,” Tuttle said. “For many real applications, you need to build a custom knowledge base and a question-answering system, and so our platform provides those pieces as well, and nearly all the other platforms don’t, so we think our platform is one of the most comprehensive, and we’re open-sourcing it so developers can learn how to build these applications.”
To date, the MindMeld platform has only been used by a few hundred developers, Tuttle said. MindMeld could face an uphill battle to gain traction as solutions like Microsoft’s Bot Framework, Kasisto’s KAI, and others have been available for years.
Tuttle said MindMeld distinguishes itself by providing advanced capabilities and the ability to handle large datasets. He also believes that, after years of experience with simpler tools, Fortune 500 companies are ready for more advanced AI tools to handle more applications. The framework may also appeal to companies that want to avoid sharing data with public cloud providers.
“It’s really designed to allow enterprises that have their own proprietary data to use the data without sharing it to the cloud and deploy it in advanced conversational applications on premise and on site, so they can use their own data and don’t have to share it with these cloud providers, which is a consequence for some customers,” he said.
MindMeld isn’t alone in its aspiration to grow adoption among enterprise customers.
In a VentureBeat exclusive earlier this year, robotic process automation company UiPath said it’s considering the creation of its own conversational AI platform. UiPath recently raised $568 million at a $7 billion valuation.