The AI Data Science Summit 2019 featured a keynote by MemSQL’s CEO, Nikita Shamgunov, where he was hosted by MemSQL partner Twingo. Nikita, a co-founder of MemSQL and the technical lead from the beginning, has shepherded MemSQL’s development toward a world where cloud, AI, and machine learning are leading trends in information technology. Now that these trends are becoming predominant, MemSQL is playing an increasing role, as Nikita discussed in his keynote. What follows is an abbreviated version of his presentation, which you can view in full here. – Ed.
Today I want to talk about the demands of AI and machine learning data infrastructure. Certainly the promise is very big, right? I couldn’t be more excited about all the innovation that’s coming in retail, in health care, in transport and logistics.
Investment into AI is very, very strong, with predictive analytics and customer analytics – areas where MemSQL is experiencing rapid and widespread adoption – as two of the top three planned uses for AI technology.
However, the data challenges remain. Only 15% of the organizations have the right architecture and the right data infrastructure for AI, and only 8% of all systems are accessible to AI workflows. And we see this all the time. You walk into a major organization, data is siloed, it’s locked into databases, SaaS services, data warehouses, and more. As a data scientist, data management becomes kind of one of the first challenges that you need to solve, because your AI programs and your AI technology are only as good as the data that is flowing in.
Databricks says the majority of AI projects have challenges moving from concept into production. What causes those delays? Typically, AI workflow is a multi-step workflow. MemSQL can simplify and accelerate a lot of the steps in the AI life cycle.
MemSQL plugs into modern applications and plugs into modern workflows, such as AI workflows, a lot better than old school technology. It allows you to close the loop and automate the loop and remove a person looking at dashboards from the workflow and make the system completely automatic. And that’s what an operational system allows you to do, so you can go from analytics to pixels, from analytics into an app, almost instantaneously, with an automatic workflow.
We are currently testing technology internally that reliably allows us to get responses to a specific set of query types in 2ms. We showed this to one customer, and they had a truly interesting response. Our customer said: “First, we actually don’t believe that you can do this. But if you can do it, we want it first.”
We’re working with top U.S. banks on fraud prevention, which is another very, very typical example of using MemSQL. And fraud needs to be detected these days in real time. You swipe a credit card and you want to reject that credit card transaction in the transaction. And so, for that, you needed to have a very performant, very efficient data backbone.
MemSQL is particularly well-suited for use in speeding up the workflows that are needed for all kinds of AI and machine learning applications. We jokingly call this a “markitecture diagram” – it shows the many ways that MemSQL brings together all the different strands of input and output, providing a fast, scalable, SQL database, which can ingest and store nearly any kind of data, for AI and machine learning programs to work against.
We hope to work with many of you on AI and machine learning applications going forward. You can see my conference presentation here. For more information, please reach out. You can download and use MemSQL for free, up to certain fairly generous limits, with community support from the MemSQL Forums. For more ambitious uses of MemSQL, or to subscribe to our excellent paid support plans, contact MemSQL today.