Tim Clarkson: Right. A little softer question. Can you explain, not the big guys, but say a smaller application, you mentioned a drugstore where they might want to use AI as their customer service. Kind of explain what that would look like or a retail shop where they’re using AI rather than necessarily people to get business done?
Jack Abuhoff: Sure. Well, I’ll give you a fresh example, not even from the work that we’re doing today, but from the work that I’m hopeful that we’ll be doing at some point in the near future. We’re in conversations with a kind of home furnishings manufacturer who wants to create the ability for someone to upload pictures to their website and to utilizing those pictures to discover which of their furnishing products would fit best within that environment, and maybe even display what that might look like. So I think as you go from enterprise-to-enterprise, firstly, I think, it’s almost inconceivable that there will be enterprises who won’t be affected and likely benefited from these technologies if they seize them correctly.
And the fact that as we do the work that we’re doing with the foundation model builders, we’re also continuing to plant seeds in enterprise and to work soup-to-nuts with enterprises to figure out how do they take advantage of these technologies and seize these opportunities is, I think, planting very strong seeds for the future.
Tim Clarkson: Right. Okay. I’m done. Thanks.
Operator: The next question comes from Dana Buska with Feltl. Please proceed.
Dana Buska: Hi, Jack.
Jack Abuhoff: Hey, Dana.
Dana Buska: Congratulations on an excellent quarter.
Jack Abuhoff: Well, thank you so much. We’re very happy with the quarter. We’re happy with how we’re kicking off 2024.
Dana Buska: Oh! Wonderful. My first question I have is that, I just want to ask a question about your Goldengate platform. It is my understanding that that’s built on the transformer architecture. And is that like the same architecture that OpenAI uses? And I was just wondering, what does that mean for your offerings?
Jack Abuhoff: Sure. So, I believe that it is the same architecture. And when we see that it is, what we mean to use that as a proof point for it is that we’re making good, solid, future-proofed engineering decisions within our engineering department. And I think that’s important, because it’s not trivial to make those decisions and it’s not obvious when you’re making them whether you’re making the right ones. Now, that having been said, we are not by any measure saying that we can use the Goldengate as a substitute for ChatGPT. That’s far from the case. Goldengate is 50 million parameters. We believe ChatGPT is 1.7 billion parameters. Goldengate does very specific things that are good for us and good for our customers in our business.
We use it in many, many of our deployments. But you can’t ask it to write a poem about butterflies and iambic pentameter. It just doesn’t work for that. The fact is, though, that, we picked the right technology. We’re using it very effectively in much of what we’re doing. It was very, very useful in the work that we were doing for Big Tech companies in classic AI. It has less utility in large language models, but continues to have lots of utility in our business.
Dana Buska: Okay. Wonderful. With the kind of fast-moving marketplace and fine-tuning and reinforcement learning, do you have any estimates about how large that market is right now?
Jack Abuhoff: I think there are a lot of different estimates. The one that we’ve shared in the past, I don’t have the data in front of me, but the one that we shared in the past was a Bloomberg estimate looking at AI in large language model-related services and showing that there would be a significant expansion in that market. I’d probably point you to that and be happy to send you a reference for that after the call.
Dana Buska: Okay. Okay. Great. That’s excellent. And in the last couple of conference calls, you talked about your white label agreement and I was just wondering, how is that going? Are you seeing any inroads with that?
Jack Abuhoff: Yeah. We’re seeing inroads. We still think it’s early days. Again, it’s early days for enterprise applications as a whole. We had a very good quarter with that customer in Q4. I think we’re going to see pickup from the white label partnership beginning in Q1 and probably through the year. But again, I view that very much as a seed for the enter — that we’ve planted for the enterprise side of the business. Right now, the growth that you’re seeing is primarily on the work that we do — the data engineering work that we’re doing for the internal builds that the hyperscalers and large tech companies are working on.
Dana Buska: Okay. And what strategies are you employing to differentiate yourselves from your competitors?
Jack Abuhoff: So, I think, it depends on the line of business. If you think about the services side of business, which is the bulk of the business, it’s 80% of the business. What we need to do is no different than any other services company would need to do. We have to do a very good job at what we’re hired to do. Just like the question Tim asked, he said, well, is the data quality really important? And I think the answer to that is, as I said, it clearly is critical. It’s what we’re being hired to do. Beyond that, you care about the level of service that you’re obtaining. You care about the qualities that the vendor is bringing to the relationship. You’re caring about how tightly aligned they are with your engineering team and whether — when they zig, you can zag and whether you can follow their lead and be responsive to their changing requirements. We’re bringing that to the table.
Dana Buska: Okay. Excellent. And do you have any new products or services that you’re excited to be introducing this year?
