Thought Leaders in Artificial Intelligence: Grant Ingersoll, CTO of Lucidworks (Part 4)

Sramana Mitra: What kind of infrastructure are your clients running these days? I imagine you’re going after the larger e-commerce sites because the smaller ones don’t have the infrastructure to run these kinds of stuff.

Grant Ingersoll: We scale pretty nicely with the size of the organization. A good chunk of the top 100 or top 250 retailers in the US are powered off of either our open source software or our commercial product. Many of them are in a migration right now, going from on-premise up to one of the big public cloud environments. We’re starting to see more of this running on Kubernetes and Docker. It really depends on the organizations and where they’re at.

Most of the really large retailers that we support have moved to public cloud. They are now choosing the one that is most competitive for them. In fact, some of them even have mandates to not go to that particular public cloud. That’s a new trend that I’ve seen starting to happen in the last year and a half or so. They’re looking for multi-region, elastic scalability, especially if they’re in the European market and they also need GDPR compliance. They’re looking for continuous uptime, elastic scalability, and multi-region fault tolerance.

Sramana Mitra: What is the adoption of deep personalization in e-commerce companies right now?

Grant Ingersoll: I would say it’s a mix. It’s definitely on everybody’s list of things they want. With all of this stuff around machine learning and all of the feedback loops, what we routinely see is a lot of people have these data silos. Marketing only sees the user. They’ve got Adobe or Google Analytics. They do their marketing-level slicing and dicing of the data. That’s where they develop their intuition around what their customers care about.

They then dump that data out to their data scientists. Their data scientists then go build some model offline using Python or R. Then they go back to marketing and say, “We just increased conversions by 10%.” Then the business says, “Great! Let’s put that in production.” Engineering says, “We don’t have Python in production here. We’re a Java shop.”

Four months later, they have that model in production. It took that long to get converted to the production language of choice. The model was flat because it was trained off a winter data. Now it’s spring. The keyword shovel in winter in Minnesota has a very different meaning than the word shovel in summer. A lot of people struggle with this. That’s what hampers the personalization. This whole data workload really struggles with that kind of workflow. We try to address that. It is often a challenge of inherent data silos within each company.

Sramana Mitra: If you were to start a new company today, what does the landscape look like? What are some open problems that need somebody to put attention on and build a company?

Grant Ingersoll: It’s a great question. We haven’t talked a lot about the enterprise side of it. People are looking for faster decision making. There’s this really interesting sweet spot that marries the core data that you care about with user behavior such that you can build up a really interesting view that helps all of that data. For instance, one of our large banking customers in New York uses that combination of their own data with all of the ways their users interact with that data.

They then feed that to this notion of next-best action whereby the sales rep or the account manager optimizes the time that they spend with each client for the level of service that that client is at. There’re some super high wealth clients. They want to make sure it’s seamless across all of that. To me, that’s an example of that really interesting sweet spot that moves AI from being about technology to AI being about humans. How do we augment things that we, as humans, aren’t very good at with technology?

This segment is part 4 in the series : Thought Leaders in Artificial Intelligence: Grant Ingersoll, CTO of Lucidworks
1 2 3 4 5