Exploring the value of business communications.
Have you ever stopped to think about how many business conversations you have in a day? A week? A month?
Expand that to include the number of conversations your colleagues have. What about emails, Slack conversations, and corporate updates? Advertisements, both print and video? There’s a lot going on, isn’t there?
Whether it’s an executive-level conference, a quarterly shareholder bulletin, a weekly standup among key stakeholders, or the many, many incoming calls each day to customer service, a tremendous amount of business information is out there, just waiting to be analyzed.
That’s why, at ainblick GmbH, we’ve developed a process that allows us to transcribe and analyze all of your business communications. Our AI algorithms enable us to dig deep into that information to find valuable, actionable insights. You can then apply those insights to fine-tune your intercompany and intracompany communication for optimal business performance.
Massive computing power
Not surprisingly, our work requires a lot of computing power. On average, we process 1,000 hours of speech for each new version. Analyzing that much data requires at least two Nvidia Tesla V100 graphics processing units (GPUs); preferably four. These GPUs are designed to accelerate AI and graphics, and each one matches the performance of about 100 central processing units (CPUs).
Just a few years ago, that would have been more computing power than ainblick, as a small, startup company, could have dreamed of using. Even today, purchasing that kind of power is price prohibitive.
Fortunately, cloud technology offers us an alternative. Today, we can tap into V100-level processing power through the cloud, paying only for the amount of time we use it. There are a lot of cloud options out there. As a startup, we’re uniquely positioned to try them all. We’ve found that IBM Cloud Virtual Private Cloud on POWER stands out for several key reasons.
Easy provisioning, unmatched performance
First of all, with IBM Cloud Virtual Private Cloud on POWER, we can provision the GPUs we need with virtually no wait—we just specify the technology we require and it’s ours for as long as we need it. That’s a pretty significant selling point in an increasingly competitive and in-demand marketplace.
IBM Virtual Private Cloud on POWER also offers a fast, cost-effective way to accomplish our tasks. In a side-by-side comparison, the IBM Cloud solution processes one epoch of our neural net training in 12 minutes (versus 27 minutes for a non-IBM setup). Even if we adjust batch size to optimize the non-IBM setup, IBM Virtual Private Cloud on POWER is still 30% faster. When you’re processing 50 or more epochs, that translates to a lot of time—and money—saved.
Getting the job done
As ainblick is still a small company, we’re in the process of developing our best practices. We’re bringing new clients on board as quickly as we can and helping them mine their business data for important new insights. IBM Virtual Private Cloud on POWER is like a secret weapon for us—a cost-effective way to obtain unprecedented processing capabilities that will get our most complex jobs done quickly.
Learn more about IBM Cloud Virtual Private Cloud on POWER.