Follow Datanami:
October 22, 2014

Graph Analytics Startup Has Roots in IBM Watson

IBM Watson’s 2011 victory on Jeopardy! was a watershed moment for artificial intelligence and ushered in a new age of cognitive computing for big data. But as powerful as Watson is with natural language processing, it’s limited in the types of data and analytics it can perform. Today, a group of former IBMers led by the founder of Watson Labs formally unveiled Cognitive Scale, a big data startup that’s building upon Watson with the power of graph analytics.

Cognitive Scale was founded 18 months ago by Matt Sanchez, who founded and led the research and development arm of IBM’s commercial Watson Solutions division, called IBM Watson Labs, following the game show win. The Austin, Texas resident worked with Watson clients in the healthcare and financial services industries before leaving in 2013 to found Cognitive Scale, which is based in that city.

IBM Watson is great at certain big data tasks, such as reading all the literature on a certain type of cancer, and then answering questions about that type of cancer. Its ability to break text down into constituent pieces and effectively “learn” the information is unparalleled in the computer industry. But while IBM is gaining traction with Watson, Sanchez and other former IBM Watson executives recognized there was a larger opportunity to harness data in different ways.

For the past year and a half, Sanchez and his R&D team at Cognitive Scale have been working on the core technology behind Insights Fabric, the name of its cognitive platform that’s hosted in the cloud. The heart of Insights Fabric is a proprietary graph database that brings together all sorts of structured and unstructured data from public, private, and social sources, and powers the applications that generate personalized recommendations and predictions.

“What it does at highest level is it sources data structured and unstructured, from different sources–including public, private, social, devices–and then run analytics on the data,” says Cognitive Scale CEO Akshay Sabhikhi, who formerly led IBM’s Smarter Care initiative. “It orchestrates different analytics on the data, and once you run the analytics it lets you compose an insight or a set of insights that gets delivered to an API on consuming applications, which could be a mobile device.”cognitive scale

The insights are highly visual, and are designed tell you four things, Sabhikhi says, including: what happened; why something happened; what you should do about it; what’s going to happen if you don’t. “So we try to bring those elements together,” he says. “Any insight could have multiple analytic services that need to be orchestrated.”

IBM Watson also plays a role in with Insights Fabric. The company is tapping into Watson for its question and answer capability. It’s also using the IBM service to create “pychographic” profiles of customers or consumers, which also boosts the accuracy of the recommendations. Also, any interactions that users have with the applications are fed back into Cognitive Scale’s system, with the aim of improving relevancy and recommendations down the road. “It’s a feedback loop,” Sabhikhi says. “The system actually learns based on those interactions.”

Shining a Light on Dark Data

Cognitive Scale is eager to help customers take advantage of so-called “dark data” that Watson can’t absorb, such as blogs, reviews, social media data, photos, and data generated from mobile apps and devices. The company procures and prepares this data behalf of its clients.

“Watson is great technology….but Watson is only good with unstructured data,” says Sabhikhi, who was Sanchez’s colleague at IBM and at Webify, which IBM acquired seven years ago. “One of the issues with a lot of analytics, including Watson, is it’s only as good as the data you bring to it. There’s a massive opportunity around the data that needs to be fed to Watson.”

The company has built two graph databases to date, including one for the travel industry that has more than 1 billion data points, and one for healthcare that has more than 600 million. It’s currently working on additional databases for the retail industry, and is just getting started with financial services. Customers don’t run the graphs themselves. Instead, they get access to the graphs, and the ability to customize it with their own data.

At a high level, the graphscognitive scale_logo show the relationships between people, places place, and things. But each database is highly customized to the specific industry. For example, in healthcare, the graph is constructed of patients, family members, doctors, diseases, drugs, side effects, cities, and employers. The graphs for travel may hold reviews from sites like Yelp, pictures from Pintrest, posts from Facebook, or tweets from Twitter.

“We have a large number of patents in how this data gets assembled, how you extract the information, how you create this a massive map of relationships, which are highly probabilistic in nature,” Sabhikhi says. “It’s not just nodes and edges, but nodes and edges with probabilities and strengths of connection and so on.”

The graphs are constantly changing as they get refreshed with live data. That’s a key differentiator of Cognitive Scale’s approach as compared to Watson’s mostly batch-oriented approach. “One of the reasons the IBM Watson team partners with us is because they see value in our platform in bringing that live-streaming type of data and the cognitive graph on top of the Watson service,” Sabhikhi says.

Early Adopters

Cognitive Scale has already worked with a handful of customers in the travel and healthcare industries, including a cancer hospital and a pediatric asthma program.

The subjects of asthma project were children who had previously been admitted to the emergency room for asthma attacks, and the goal was to give healthcare workers the right information to keep them from returning to the ER. The graph generated for the client included basic data, such as patient histories and EMR data. But Cognitive Scale also procured data about pollen counts and ZIP codes, which it uses to generate recommendations for specific people to stay indoors at certain times.

On the travel front, one of Cognitive Scale’s first customers is WayBlazer, a new travel website co-founded by Terry Jones, who founded Travelocity and was a founding chairman at, and Manoj Saxena, who is Cognitive Scale’s chairman and formerly was general manager of IBM’s Watson Group.

WayBlazer uses Cognitive Scale’s cloud to provide a personalized travel concierge services. It generates recommendations by linking together the different places, offers, and preferences with customer’s social, cultural and economic data, and packages it all up with a Watson-powered natural language interface.

This co-mingling of deep analytics with real-time data and personalized recommendations is on the cutting edge of big data today, and will likely become more common in the future. “Some of the use cases we’re going after require live streaming information,” Sabhikhi says. “I need live streaming information that’s constantly changing that’s updating relationship in my graph that need to surface real-time insights to my customers.”

Related Items:

Retailers Embrace Graph Databases

IBM Makes A $1-Billion Bet To Make Watson A Business

IBM Makes Watson Technology Available