The Exchanges

Q And, uh, so you have all of that, and then you, um, funnel it, uh, into the platform, or I guess, I guess maybe you do all of this once it's in, into the platform already? And you said the platform is, uh, is Google for, for that, for, um.

A Yeah, ZDP is built on Google, BigQuery right now. So, you know, all these pipes are that are continuously pushing the data to the platform, or sometimes we are pulling the data, then you have to definitely store it on, on a platform like that, and then you go to the processing on top of it, right? So you have to sometimes join these, this data, extract certain attributes for each of these, create a graph, because these are related to that, Uh, related to each other, you know, a person works in a company, or this company maybe sells a product to this another company. All these connections are there that we are extracting out of that, and that is done through, you know, different technologies, you know, either spark code or data flow that GCP has, and all kinds of basic processing. Machine learning, definitely, right? Model building, some of these are just traditional machine learning. Let's say you're gonna have a, Tree-based model that, you know, use that data to create, you know, classification or prediction. Also Gen AI, right, using that data to either create an additional context for Gen AI, also just to make sure that you can prevent almost hallucinations saying that, you know, do not go beyond that. Here is your data and try to extract insights out of that. So all that processing is done over there. Then you gotta push this tool.

AI assessment note: “Yeah, ZDP is built on Google, BigQuery right now.”

Q What does the engineering organization look like today in terms of size, um, you know, across, uh, I don't know, we'll, we'll get into how you organize, but like across engineering, data, AI, what's the total number of people?

A It's roughly, you know, we're at 900 people, close to a thousand people, uh, So, um, the way we are organized, maybe we're gonna come back to it like you're saying, uh, definitely we have a strong platform team, and data platform is part of that. Data platform covers data engineering, data science, and data platform by itself. Uh, there's a separate enterprise engineering team, we call it enterprise productivity team, because we care about productivity a lot, and that deals with, you know, serving the internal customers revenue team, finance team, HR team, and all that. Uh, definite operations team, you know, site reliability, um, security teams, IT teams, and on top of the platform, we have, uh, multiple products. Uh, these product teams, you mentioned sales product, marketing product, operations, talent. We have a chorus for conversational intelligence, uh, and a couple of more.

AI assessment note: “we're at 900 people, close to a thousand people”

Q forward to the conversation because it's talking data, Uh, with somebody who runs data at a data company, so it's, it's like, it's, um, you know, it's like very meta in the, in, in, in some ways. So, uh, to start from the start, I guess, uh, there is a core data platform, which is a foundation for everything. Uh, is that correct? And if so, how does that work?

A So the, the way it is structured is definitely we have a cloud infrastructure team underneath. We are using both, uh, Google Cloud as well as Amazon Cloud. On top of it, uh, we have actually two data platforms. One is, um, for, uh, for all the data that customers, you know, need and our products need, and there's another one that's for internally focused for our internal customers, and they're built on different technologies. Uh, there's a, you know, pipe between them just to make sure that data is exchanged in real time. Um, then you, we have, uh, teams that are developing applications on top of these data platforms, uh, including data science applications, you know, creating machine learning models, using Gen AI, you know, using that data. Um, and the fact that we have the data in one place, uh, makes it super, super, uh, easy for, for, to do all of that. And also, you know, data is huge in our case, like we are, we are the market leader in that with respect to company data, you know, intent data, you know, Contact data, first-party data, third-party, other, other third-party data. Um, then we have on top of it, uh, what we call, uh, product foundations teams. For example, search team, identity team, you know, um, um, admin, admin hub team that, you know, you deal with settings, permissions, and all that. Um, um, platform UI team, for example, and all the teams are, uh, we ca…

AI assessment note: “we have actually two data platforms. One is, um, for, uh, for all the data”

Q And then, uh, other than, uh, my internal data as a customer, do you presumably have, um, access to more private proprietary kind of data sources as well as part of the platform or you only do, uh, publicly available?

A Publicly available, but if customers, definitely first party data, right? So some costs, Companies, um, let's say, you know, they're using co-pilot. As a result of that, they want to actually combine their first party data with our data. Let's say, you know, they have a set of, um, uh, contacts, right, in these companies, and the profiles for each of these contacts that they have may not be complete. So they want to use, let's say, our, uh, our data and tech to enrich that, to add, you know, either to correct these, to remove duplicates and all that, but on top of it. Uh, maybe to get additional attributes that they can use for that purpose. So for that they can, we can get that data on their behalf, uh, completely, you know, dedicated to them, uh, for as a single tenant. And, uh, we combine that so we can extract additional signals for them for, for selling.

AI assessment note: “Publicly available, but if customers, definitely first party data, right?”

Q So you do all the things as an, uh, application, but as I was, um, prepping for this, you, you also have a data as a service business. I don't know how big, uh, each one is, but what, what's that part, the data as a service business?

A Yeah, data as a service is also, uh, we call it DAS is one of those businesses. So some companies, um, even though they are customers of ourselves, they have licenses, you know, seats, That their sellers teams are using our product, but sometimes let's say you wanna, they wanna get a copy of that data and, um, enrich within their own platform. Let's say they have their own, uh, first party data. Our platform enables you to bring first party data. We have lots of third party data. We can actually combine this on your behalf very easily, but at the same time, some companies prefer that. And, uh, as a result of that, you can just get a copy of this data through multiple platforms. You can go to, you know, AWS, uh, Databricks, Um, Snowflake, you know, it's ready to go for you. It's in the marketplaces. So it's an API and you can get through API, you can get through as a file, or you can get, it is ready to go in the marketplaces for these companies, Databricks and Snowflake and GCP and AWS, and it's ready to merge for you. Uh, basically you can just do that processing over there. Um, so either way, basically my point is, I mean, our point is not to create any friction. Uh, point is to help you to make the sale, to make your sellers best sellers, salespeople. And as a result, if you need the data, we also have a capability like that.

AI assessment note: “wanna get a copy of that data and, um, enrich within their own platform.”

Q and all the things. So, presumably, it ingests massive amounts, not presumably, that's well documented, but it ingests like billions and billions of data points. Yeah. Uh, into it. So how, how does that part work? Where do you get data from all the sources? And, um, how does that, uh, get, uh, selected, funneled into the platform, cleaned or cleansed? Um, you know, walk us through the whole pipeline.

A I mean, with respect to, let's say, you're gonna get the first-party data. Uh, if first-party data is, let's say, in Salesforce or HubSpot and, you know, Platforms like that. Then we have already these pipes set up for you, so that's one way for us to get data. Uh, with respect to third-party data, we are definitely continuously, uh, crawling company websites, right, uh, merger and acquisition data, financial data, SEC data, you know, all kinds of information that are about people, uh, including news that's coming out of those, uh, that's also continuously happening. Uh, so all the public sources, basically, we have, we will have it. Uh, we have contributor network, right? Uh, for example, if you are using, uh, zooming for light, then as part of that free access, uh, we, we get some data from, from our customers.

AI assessment note: “with respect to third-party data, we are definitely continuously, uh, crawling company websites”

Q You are, ok. So your LLAMA runs on Oracle GPU? Ok. Interesting.

A Exactly. Then we can run it at any scale, because we already paid money for it, right, and for those use cases, uh, Even if, right, it is, it's comparable for, for most of the, you know, uh, um, experiments that we have done, but let's say even, even if it's just a little bit below in terms of accuracy, but that, that scale, the advantage that you are getting is still a big magnifier to whatever, whatever, whatever else you can do right now. If you do not have that technology like that, what are you gonna do? You have to, Crawl the data, extract all those insights, right? Create all those models. Maybe you have it. If you don't have it, you have to create them from scratch, do feature extraction, right? All that, all that time saving is given to you by, by using that, so.

AI assessment note: “Exactly. Then we can run it at any scale, because we already paid money”

Q So you got the data from the sources. It's all in the data warehouse. Um, we talked about, um, the, the, the sort of the ecosystem around the warehouse and the various tools you use. And then the next step, uh, after that is that you push the data to the applications. Um, and again, the applications can be SalesOS, MarketingOS, those are the applications?

A Those applications are using additional, uh, platform components, for example, search. If you go to SalesOS, there's lots of filters that, you know, companies of that size, that revenue, people of that title, and all, I mean, intent of those categories. So you do that search, which means the data that we just extracted should be pushed to search indexes. Right. Um, um, so we are using solar internally. There's also some elastic used for different purposes. Um, if you are going to join the data, then they go through, through what we call entity resolution. So there's a separate pipeline for that one. You cannot just put from a regular join, you know, as you know, that, that join is pretty involved and lots of also machine learning and rules, um, rule engines are used for that. Um, so that's one place, for example, let's say you need access to, um, You know, you find what you're looking for through search, but you need more data about the key that you just found out. Then we have a separate system to give you additional information about, about that entity, whatever that is as a company or contact, right? So all the right places it has to go. Then, um, through the APIs, our products are accessing those. So it's also unified in that sense. So there's no need for creating, um, a separate search for talent versus, you know, uh, Versus sales, because it's a big investment. It's a ver…

AI assessment note: “Those applications are using additional, uh, platform components, for example, search.”

Q with startups versus Working with large companies and in, in large companies at this stage, I would probably include OpenAI. Um, what I'm talking about is like the small, you know, Series A, Series B startup, uh, providing, um, whatever, you know, LLM evaluation kind of thing. Is that, is that something that you would consider or you basically a company like you would wait until GCP has the feature?

A Yeah. I mean, we are, we're, you know, very technical ourselves, so we can evaluate them, I think, pretty well. Uh, we work with so many, so many vendors, small or large. The point is, you know, how it is helping me. Uh, am I saving? Definitely time. Productivity is top of the mind for us. Uh, as well as, uh, if I were to do this ourselves, right, you know, it, you know, we can do anything. In the end, it's a software, but, you know, how many people you are going to locate, how much time it's going to take, and all that. So if there's a solution out there that we can use, we will do that. Uh, but one key is it has to be very easy to integrate, and it should work. Right. And we were going to make sure that through evaluation, that's the case.

AI assessment note: “we work with so many, so many vendors, small or large.”

Q So you use your data as a, basically as a rag, kind of, um, kind of, uh, setup?

A Yeah, let's say, you know, let's say account AI, uh, use case. Um, so how am I, how am I going to extract that account information, right? So this is my account. Let's say I'm the seller. Right, uh, what's this, what are the pain points of this customer? Right, what are, what were the, some, you know, how many conversations have happened so far, and in each conversation, what was the summary of the conversation? Was there a need mentioned, or competitor mentioned, or, you know, uh, there was a problem that was mentioned that I can pay attention to? I know exactly what kind of, um, discussions happened about their problem, uh, needs, as well as, Pain points, and what kind of responses we have given, right? All that has to be extracted, and this is going to be extracted with not only the information that we have about this account, but also the first part of the data, let's say you're going to bring, because let's say you were using Chorus, and all that, you know, you own that data. Yes, our technology is recording it on your behalf, but you own that data, and we extracted this on your behalf, um, automatically. So, you know, it's in one place, so it's just one, one, one use, use case of, uh, Copilot. Um, the others are, Actual signals themselves, right? So it's going to tell you, let's say you work with, um, you know, John Smith in this company and the person has moved to anothe…

AI assessment note: “Yeah, let's say, you know, let's say account AI, uh, use case.”

Q So you crawl all of that, uh, you combine it with some internal data. Um, what happens then? Like you just, uh, how do you clean and normalize everything?

A Depends on type of the data, definitely. Uh, if you are, let's say, you know, getting public, uh, public, uh, public data from public sources, um, first of all, you gotta crawl it, right? You gotta crawl it at a certain frequency so that you are gonna get, um, you know, all the changes that are relevant for our purposes. Uh, you have to put the right tech over there so that the data that you are extracting from those sites will be correct, right? Uh, we also have a good number of researchers about, 300 of them. If there are specific cases that we need a human, human touch, you know, for them to actually correct the data, it's very difficult to find, uh, definitely in certain cases to verify that this is accurate data. And also, uh, people might be involved in the loop as part of that.

AI assessment note: “300 of them. If there are specific cases that we need a human, human touch”

Q And is that because the earlier model, uh, still works better than Genative AI for that specific task?

A With respect to extracting from, uh, we have, we have not done it. Actually tried it in that sense, you know, extracting from recorded call. Um, um, There were some experiments on that. I think it's going to be very expensive to run this unless we have our own model now we are running on, uh, GPUs that we, we have created. Probably this could be one of the, one of the areas that we can do, but since the, our NLP pack for creating transcripts is working really well compared to, you know, uh, existing, uh, what's, what's out there. Um, it's not the first priority to replace it at this point. Um, but any ups, Gen AI was the, you know, that was the first use case and wow moment. Um, then with Copilot, as you know, right.

AI assessment note: “it's not the first priority to replace it at this point.”

Q What, what has been most challenging? What has maybe not worked and you went in one direction and then you're going in another direction? Like any lessons learned that may be interesting for people to, to hear about?

A I think across, so creating a data platform with respect to its components, right? All the integrations on the input side, output, Put side, storage, processing, you know, governance layer, uh, which is like the cataloging, you know, security, uh, permissions, um, definitely pushing the data for business, uh, intelligence, you know, the Tableau and systems like that. So most of these are well-known, but just gets learned and relearned every time, right? So I'm lucky that I have created, you know, huge platforms multiple times in the past, so it's, it's, it's, it's, it's, it helps, yeah. It helps, but, uh, you know, choosing the, Choosing the, those components, make sure, making sure that data flows, let's say, in real-time streaming mode, you can be assured of the quality of that data. Uh, if you're going to create, let's say, analytics out of those BI systems or custom analytics interfaces, right? All of these should be easy. I think the lesson is, um, um, this, again, that, that gets learned every time is, uh, you have to pay attention to All these pipes and, you know, the boxes and how these things are connected to each other so that you can do it in a way that you can trust the data. You can turn around and, you know, time to market is very important over here. You don't have to start everything from scratch. Um, let's say you're going to do analytics. There are lots of ana…

AI assessment note: “You don't have to start everything from scratch.”