Verizon >> IoT Cloud

With over 140 million subscribers, Verizon was faced with three challenges in bringing their consumer IoT platform to market:

• How to create a cloud system that would scale to the size of their user base.
• How to interoperate with their existing network and broadband technologies.
• How to build a system that would be flexible enough to deploy consumer IoT services defined in the future.

To overcome this challenge, Verizon's IoT service was built on SiliconMint's stream processing platform, helping it scale to tens of millions of homes.

  • The Challenge
    Is it possible to build a consumer IoT cloud that scales to over a hundred million subscribers, without adversely impacting existing hardware, networks and services?
  • The Challenge
    Can human and machine intellect be combined in a soft real-time system to provide meaningful benefits in image and text processing tasks?

Yahoo Japan >> Human + Artificial Intelligence

Faced with increasing demands on categorizing and classifying the content of ads in real-time, Yahoo Japan faced a unique challenge:

• How can the tasks be done in near real-time without increasing the size of their moderation workforce?
• How can the system guarantee both accuracy and precision of the resulting work?
• How can new employees be onboarded quickly, across a diverse range of skill sets and backgrounds?

A system that combined artificial intelligence with the power of the human mind was built, such that Yahoo Japan could categorize and classify information without increasing the size of their workforce.

eTrade Competitor >> 1 million events / second

Backed by Goldman Sachs, a broker dealer was winning mindshare in the marketplace, but faced a set of technical challenges to deliver key features:

• A way to provide real-time stock quotes for the purchase of complex baskets of securities
• A way to detect and alarm on stock price information in real-time
• A way to execute automated trades based on what the market was doing in real-time

With SiliconMint's stream processor, a system that could withstand the 1 million events per second at market peak was built, in addition to mining this information in real-time for alerting end-users and executing automated trades.

  • The Challenge
    Is it possible to both process and intelligently act upon 1 million stock price events per second
    in real-time?
  • The Challenge
    Is it possible to let scientists iteratively create and deploy voice-enabled applications in the span of a day?

Government Research >> Iterative Voice Application Generator

A leading team of research scientists created a voice recognition engine that rivaled the accuracy of Apple's Siri technology. Their next challenge was to bring the solution to market. They needed:

• A rapid way to generate a myriad of mobile applications across a diverse set of verticals.
• A low latency mechanism to bring voice from an end-user to their cloud-hosted engine.
• A mechanism to gather metrics in and drive generated application behavior in real-time.

To meet this challenge, a system was built that allowed scientists to generate mobile applications without code, deploy the applications to a specialized cloud environment enabled for voice, and to gather real-time metrics on application behavior to iteratively define the next set of experiments.