46:29 Webinar
See Everpure Data Stream in Action: Stop Prepping and Start Shipping AI
In this session, Everpure experts will demo the new Everpure Data Stream service—our purpose-built answer to NVIDIA’s AI Data Platform and the “data prep problem”—by building an AI data pipeline in real time.
This webinar first aired on July 28, 2026
The first 5 minute(s) of our recorded Webinars are open; however, if you are enjoying them, we’ll ask for a little information to finish watching.
Click to View Transcript
Hello, everyone, and welcome to our expert-led demo, Stop Prepping, Start Shipping AI. Over the next forty-five minutes, we're going to do something a little bit different. Instead of telling you about Everpeer Data Stream, we're going to build a live AI Data Stream pipeline with it in real time in front of you.
And I'm so pleased to be joined today by Austin Boone. Austin, welcome. Hey, Erin. How you doing? Doing great. Doing great. It's so good to have you on here today making your debut as part of our webinars.
Thank you so much. Oh, thank you. It's a pleasure to be here. Yeah. Austin is one of our field solution architects. So he spends every day with real practitioners building AI solutions in the real world, which makes him exactly the right person, to be showing you, everything that Data Stream
can do today. All right, so before we get into it, and before we actually get building, Austin, you've been in the field with customers since joining Everpeer. What is the most interesting thing that you've learned about where AI actually stands in the real world today?
Yeah. So what's awesome is that we're largely in the beginning stages of experimentation, right? Many customers have even started to deploy their own AI workloads, but still, the technology's advancing so rapidly that being able to optimize and continue to glean, increasing levels of value from AI and data analytics and all of those different,
things is still growing. You know, organizations are in, a cycle of maturity. Mm-hmm. So as they implement different technologies, they realize they need to go back and address governance, data preparedness, to get increasing levels of value from that technology.
So it's an exciting time, to be a field solution architect for AI and walk with customers throughout that process. Yeah. Oh, that's so, that's so exciting, to get to be a part of all of that innovation right now. You mentioned experimentation, and I think the thing I continue to hear from you and other
FSAs is that, and really the research shows this too, is that we are at this precipice where we're moving from that experimentation to actually bringing AI to life in the enterprise at scale, right? There, there have been a re- lot of really promising proof of concepts built, and some really promising first forays into actually, putting AI into production.
But largely in a lot of enterprises, this is still in experimentation mode. And trying to get out of experimentation mode, as we'll talk about, and actually succeed in production largely depends on the data, right? So, um- Absolutely it's, it's interesting to come around full circle to this moment, and everything comes back to data.
So, yes. Fantastic. All right. So, with that, let's talk a little bit, about where we're headed today. The plan for the next forty-five minutes, is really to, to start with the data readiness problem, talk about why so many AI projects are stalling
before they ever hit production, and why the bottleneck really is almost never the model. We'll keep this part tight, I promise, and then we'll be onto the main event, where Austin is going to build a live AI Data pipeline with Everpeer Data Stream end to end. We'll zoom back out to show how Data Stream After that, we'll zoom back out to show how Data Stream fits inside the larger NVIDIA AI Data Stream platform design, and then we'll close
with your questions. So please do make sure, there's a Q&A, there's a Q&A, panel that you can put your questions in. I'll be monitoring that. We are live right now. And, I will answer questions throughout, and then we'll leave time, at the end as well. By the end of this session, our hope is that you will have seen a repeatable automated path
from raw data to a production-ready AI pipeline, and, and see, see the, the possibilities for what that can, can do for, for your AI efforts. All right. Before Austin starts building, let's spend just a couple minutes on why this matters. So what we see in the research, is a few different things.
These stats come from our, s- our joint study that we did with IDC earlier this year, and there's a couple things that, that make this portion of getting data ready for AI so critical. First of all, 90% of enterprise data is unstructured, which means that when it comes to AI, if we are only giving AI structured data, then we're missing out on
a, a majority of the information, that we have available to feed AI and to get intelligence from. The second thing is that right now, most AI proof of concepts never reach production, which is, you know, that's what a proof of concept is for, right? But there's a lot of investment going into AI that is never actually delivering
value to the business. And a lot of that, based on this, this research, the indication that we had is that a lot of that, that breakdown is from data, and data actually not being ready for AI. And I'm curious, Austin, does this ring true for you in the projects that you're working on
with customers? Do, do these stats sort of, sort of hold? Yeah, absolutely. And that goes a little bit about, It goes back into what we were talking about before, kind of like that experimentation in the form of proof of concept, proof of value. So after you get past, hey, do we wanna, build in the cloud?
Do we wanna build in the on-prem environment? Which GPUs are we gonna leverage? Which models do we want to leverage? To really which problems are we trying to solve, right? So that's why a lot of those proofs of concepts, they are built in kind of a, a lab
environment, may not be right size or extensible into a more enterprise application that supports various functions across an organization. So being able to, take those lessons learned from the proofs of value, and then, inculcate those into future iterations of proofs of value, that are able to scale in a well-governed environment is important.
Yeah. Yeah, absolutely. And, and, I think what you just said, anyone who has worked in IT, understands this very well. You can create a great proof of concept, with, with fake data and with, you know, an ideal environment, but then when you go to bring it into production, things like data quality, things like, the infrastructure you're
actually deploying on, right, all of those things end up mattering so much more. And, and in some ways are, are, potentially harder in terms of bringing AI to life. And again, as anyone who's been in IT or ever tried to create an ATL process, build a database, or build an AI data pipeline today knows, it's not simple, right? It takes, it takes a sophisticative te- sophisticated team of people, and these are
folks who have, really high-value skillsets, like data engineers, data scientists, ML engineers, or MLOps developers, right? It takes those teams months to create a da- a- AI data pipelines today without some kind of assistance from, from technology. And those folks could be building the next, AI application.
Those folks could be, those folks could be figuring out the next application for, for machine learning, in, in their organization. And so, what we, what we love to see here is that, that w- we are hoping that we can help free up those really valuable resources to work on higher value projects than simply getting the, the data ready for, for AI.
And- And Erin, just- Yeah, go ahead just to dovetail on that, what you just mentioned, I've, in a previous life, held multiple of these hats, and many times we actually wanna get towards the business side of data engineering or solving problems, using data science, but we are always constantly trying to stitch together different pieces of software and technology, building justifications to even acquire the right tools
that we need in order to build those data pipelines. So it's, it's sometimes it's like those non, you know, specific, things that are related to our own practice that cause us the most pain, right? And it's really, you know, going back to, like, the IT portion when you wanna get into the data and start to produce value more rapidly.
Yeah. Yeah. I think I've, I've seen, I've seen varying stats from anywhere between 50% of data scientist time to 80% of data scientist time that is spent on actually just getting the data ready, right? And, and like you said, there's, there's so much potential for, for what you can do when you're in one of these roles, that, it can be a little, a little demoralizing,
I think, sometimes. And I experienced this as a data analyst back in the day, so I can relate to what you're saying, Austin. When you're spending most of your time just getting the data ready, right, and not actually getting to spend your time doing something meaningful- Mm-hmm with that data. So, yeah, that's Couldn't, couldn't have put it, put it any better.
That's fantastic. And yeah, we, we believe at Everpure that there is a better way to do this. NVIDIA also, believes that, which is why they created the AI data platform reference design that this is built on top of. And, and, that's, that's exactly what we're, we're hoping to bring to life,
for everyone listening in who, maybe is a little bit tired of, of creating data pipelines. All right. So with that, let's get to building. Austin, are you ready? I am. I am.
Fantastic. So- This is the, this is the fun part. Absolutely. All right. Good stuff. So, what we have here is our user interface for Everpure Data Stream.
And just to orient everyone to the screen, what you would see here in the middle is kind of like, a data catalog, data pipeline catalog, and what we're calling kind of like our streams catalog view, where you can see the streams, that are available here, the ones that are published, streams that are still in draft. You can see up at the top right-hand corner, you can actually create a new stream.
And then on the left-hand side, what you can see here is there's the stream catalog. There's a built-in semantic search capability that allows you to search across various streams inside of Data Stream in order to test functionality. You could do the same thing with agentic chat. You can toggle between different, functionalities because Data Stream actually
does provide large language models, various versions of that, that are optimized, inside of our platform based on your infrastructure. In addition to that, there's a functionality here that allows you to create users and provide fine grain access control. You can grant access.
Let's just say you are an administrator, and you've gotten, a request to build a stream or a request to grant access to a stream. You can, see whether there's a pending request, who the owners are, what their permissions are, who's approved, what's been rejected, etc., in order to provide a little bit of granular platform level access to the streams that are created.
There's the ability to get help, in a form of documentation, and then generating API keys. So what we'll do here is we'll jump back into the streams. I've got a particular use case that I've been working on related to, legal. I've already created this stream, and once you go in here, you can see, there's a little bit of a description here.
The stream was created for AI-ready data supporting legal ops. You can see, that there's five hundred and ninety documents there. You can see that was last updated, a day ago. You can see your owners here, who the readers are. Some of these entities could be individual people.
Some of these could be, service level accounts that have access to read from this stream. And then we also give you the ability to manage users right here within the stream, and grant various levels of access. So Austin, I just wanna pause here, and just make sure that everyone here is, is, understanding, you know, kind of, how much is being shown here.
This is really powerful stuff. So the stream itself, that is basically an AI data pipeline, right? Absolutely. And so all of, all of those that you saw in this demo environment are, are AI data pipelines or streams that have been created by different users potentially.
And now this console, you can have someone actually managing and governing this, right? So you have access level controls at the users and access requests. So you have governance built in at the data layer. And then those search and chat capabilities, I think this is You know, when I was first seeing these demos, I was, I was sort of like, "Oh, okay, that's nice, but what does that
really do?" Right? This is AI on top of AI data pipeline building, right? So this actually allows you to interact with all the streams that have been created, and search for the information you need, but also search the underlying data, that has been connected through Data Stream, and be able to ask questions of it.
Is, is that accurate? Am I understanding this correctly? Absolutely. Absolutely. You're, you're spot on, Erin. And, just to go a little bit further, an individual stream is a collection of curated data sources. So, each stream is a pipeline, and what you'll see here, I, I'll kind of show you here in
real time. What you'll see here, towards the middle of the screen under data sources is I've got, four different data sources. I've got an NFS data source that's pulling in some legal documentation. I've got an S3, data source that is connecting to, an S3 bucket that's been, generated for a specific type of legal data.
I've got another S3 bucket that we wanted to pull into this stream that's pulling in, some text files, some, some various, transcripts, etc., related to legal proceedings. And then, to simulate some file shares within the organization, pulling information from different products that are being developed, supporting legal ops. I've got another, NFS, ingest that I've done, into this stream.
So what happens is as each one of these streams are run, there's an ingest pipeline that is designed to leverage certain rules. So, for the sake of this demo, we've got two models. We've got, a text embeddings model, and we've got, a vision language model here, that are supporting the embeddings for each one of those data sources that we configured on the
previous screen. In addition to that, there's some curation rules that allow us to provide, you know, a level of, redaction or omission of certain types of data d- while, the index process is going on, throughout the pipeline. That could be, either, based off of word count, it could be off of dirty words.
In addition to that, we've got some classification capability, where we can identify different, tagging and labeling criteria for data as it's being processed throughout the pipeline. So going back to these data sources, what you'll see is each one of these data sources, becomes RTO of that collection.
It is all, vectorized. All the embeddings are based off of the models that we've identified here in the configuration, and then all of the different curation and classification rules are applied to this data based off of our data strategy. This is really an initial, capability all within this user interface in order to provide,
an orchestrated way to create, a pipeline of data that can be accessed by higher level applications, whether they're agentic, systems, whether they're apps that you've built that are gonna consume this data and- and then you can actually have lineage and understand exactly how and where that data is coming from. Very, very cool.
So looking at this, I'm just thinking back to that, that slide I had, you know, identify and discover one to two weeks, ingest two to four weeks, curate three to eight weeks, so on and so forth. What steps did you just cover in this? This was identify and discover, and ingest. Is there a curation that happened here as well?
Absolutely. Curation happened, and then also, one thing we kind of, leave off is actually serving, right? And consuming, the data. I'll show you that here in a, in an application that I've built on top of Datastream that is actually gonna leverage, the large language model inside of Datastream,
some of the semantic search capabilities, connecting all via API. So let's go over here. I basically built, an application as a proof of value for a customer. This application is, uniquely focused on courts intelligence and, specifically traffic citations.
So just to orient you to the app to kind of show for demo purposes, right here we've got, verification that we're connected to Datastream. We've got verification that we've connected to that court stream that I created a few moments ago, and that the connection to that court stream is live. If we were to go- So- Go ahead on this application, just if, you know, make sure I'm
understanding, you're essentially building the, an application that can help, citizens self-serve on understanding what is going on with their traffic citations. Absolutely. And all of that AI data pipeline that you built to power this took minutes to make. Is that right? Absolutely.
If we were to kind of go back here to the, the ingest, you can see, on this schedule I ingested, 201 documents. It was, two point, about two and a half files per second, that we were getting some pretty good speeds. But, long story short, all of this documentation was purposely curated in order to feed a citizen, citizen-facing application
that helps people, find out pertinent information related to their court cases. Can I ask what you built the application itself on? Did you use, like, Claude Code? Yeah. And actually I did use Claude Code. That is a fantastic tool that's been allowing us to build, proofs of value more rapidly.
Yeah. It's also a demonstration to, some of our customers that whatever tools you use to build applications, even if it's Claude Code, you're able to, build compatible app- apps with Datastream. You don't need anything, you know, more robust than what you're currently using. What is interesting to me about watching this, and I, and I hadn't thought about this before
when watching the Datastream demos, is Datastream is kind of like Claude Code but for data teams, right? Mm-hmm. So there's so much you can do with Claude Code. And you know, you don't even have to be a developer to, to build an application. But of course, at least for me, I've, I've always been able to, to, you know, write a
decent, application in a test environment. But again, when you go to deploy it, then you have to think about the data, the environment you're deploying. You know, I know that there's other tools out there that are taking care of that, that deployment, aspect and, and automating that piece.
And, but this is sort of the equivalent of a Claude, of a Claude Code, but for data teams, which, is, is really powerful. Absolutely. Absolutely. And provides that single pane of glass- Yeah with that unified view of the governance of each one of the streams.
O- a single place to go back and troubleshoot the streams, right? Like sometimes- Yeah data may become stale, you don't know where to go look. You can check your stream, and then go and, check your upstream data sources. But it provides that, that, that integration point for you to actually be able to troubleshoot and provide better value. Yeah. Yeah.
That's fantastic. All right. So I just wanna show, the audience a few things here. So I, obviously, we're all green here, but what you'll see in this Datastream platform connection, these are all the, the standard, open API, calls going into, authentication.
I've authenticated as a specific user with access to this stream. You can see here that, I'm connected to the models that are inside of Datastream right now, in our demo environment to support the, the chat and conversation capability. You can also see the, the, the streams API that provides the, the streams that I actually have access to, but because I'm leveraging this a- particular
application, the stream that's, associated with this app is the court stream. And then we've got a search API that is providing semantic search capability for the app that we built. So just to jump in a little bit, let's just say I'm, I'm, I've gotten a, a speeding ticket. It's $35, but when I got to the clerk's office,
they say I needed, 238 bucks. So what you see happening right here is the call went out to Datastream over the API- And let's look at this. Went over the API. It's starting to retrieve directly from the stream, right? So it's going back into, the index, and it's pulling this grou- these grounded answers back
and providing, a rich set of insights for this user. So now they can see, leveraging the large language model and those grounded answers, hey, based on documentation, you have a 4.8 multiplier. Here's all the additional fees that are associated with that, and then here's the bottom line. You can also see here we've ac- been actually
able to pull, via the NFS connection, an actual raw document. You can see some tags here that I wanted to pull in to just kind of show some of the power of the tagging and classification, some of the metadata that we're able to pull in. And then in addition to that, you can see a text view. This text is actually being pulled directly from the index, that was ingested from this
raw PDF. So, depending on how you're building your application, you can use the organic capability within Datastream in order to optimize that experience, and then also manage the way that data are being presented back within, your customer environment. That is very cool. Yeah.
That is very cool. Yeah. And, another thing I'd like to show you, again, this app was really built around showcasing, what Datastream is providing, not so much the app itself. The app is just being fueled by Datastream. It's being fueled by the large language models, the semantic search, the, microservices that
are running, in the cluster, providing all of these services. So here, let's look and see about some of the, the semantic search capability and some of the agentic, features. So right now, I wanna compare two cases. So I wanna look at these cases.
I'm comparing. I'm gonna get, really a deterministic, output initially, right? So it was able to reach back into the stream, provide information about two different cases, give me a de- deterministic, like, view of, some comparisons. But I wanna add, an AI summary to this.
So what's happening, I'm going and I'm clicking the, AI summary, and now a call went back to Datastream to Nemotron, generated this output, and then provided it back so that I can get increasing levels of value from the data that I'm pulling into my application. And it's all using the, the integrated capabilities that exist within Datastream.
Very cool. Yeah. And Austin, how many data sources were in this stream? I, I, you know, wa- looking through this, this is, this is so much, so much information to be able to surface up in the application. And then again, I'm just thinking about what it used to take to build an application like
this, versus what it takes now, with, with the power of, of AI and, and with the power of Datastream. How many, how many data sources are, are driving into this? Yeah. So right now in this stream, we've got four data sources.
Four d- four separate corpus. We did that on purpose because the particular customer that we were showcasing this capability for had several different vectors that data were coming from in order to feed their legal operations. Okay. So each one of these is, could be a court
proceedings, it could be a citations database. It could be, you know, you name it. But all four of those were ingested in order to, be able to provide those curation rules in order to sanitize data, and present data that was suitable for public consumption without any, kind of leakage of inappropriate data, for the purposes
of the app. And it's a mix of structured and unstructured, right? So the citations database conceivably, has, is structured or has structured elements, and then the PDFs themselves, are unstructured data. Is that fair to say it's a mix?
Yeah, absolutely. That's a mix. So in inside these corpus, in each one of these data sources, is a stru- a sem- semi-structured and structured data, just to, be very fair about what's in the corpus. Yep. Okay, great. All right. So, just to show you a little bit more power, if we, had a requirement to add an
additional data source. Everybody loves the app. New features are, are developed for the app, but it requires additional data, to be processed into the pipeline. So we can come back to this stream.
We can add a data source. So let's just say I'm just gonna create a test data source. I want it to be NFS. I'm gonna go ahead and connect to, a repository. Does not have to be, running on Everpure.
This can be anywhere in your data estate that is accessible. I'm gonna go ahead and load those directories. I'm gonna select a directory. I've gotten access, right? An indicator that we've got a successful connection, and I'm gonna mount that directory.
So what's happening now is it's going across the, the environment, identifying, the areas that I have access to and the different mount points. I'm gonna mount a specific path. And now I see, hey, I need to add some additional information from, the disk test database, right?
So I'm gonna select that I can identify which types of documents I want to be included in the ingest. And then I also wanna say how often I want that to happen. This is also something that can be, tailored to your unique requirement. But right here we've got hourly, daily, weekly.
So if you have, you know, kinda batch jobs that you're used to running, you can actually synchronize Data Stream for whatever other processes you have going on, externally, to Data Stream in order to consume that data once it's fresh, right? And then you can go ahead and create that data source. Once that happens, it gets queued, and then the process starts again.
As it's- And now when you go to your application, you have access to that data. Is that accurate? Absolutely. As soon as the process finishes queuing, and the, the pipeline is complete, you can begin to access that data directly within whatever application, or enterprise system that you are making that data available to.
Okay. There's two things that are really standing out to me watching that portion, and, and something you said, said earlier too. It's the flexibility and the interoperability, right? So there are other solutions out there that let you do a piece of this, right? So maybe they let you do this, but only with
the data that's on their system, right? Mm-hmm. So we, we actually allow you to, to reach out to systems that aren't on Everpure's platform, right? So there's that, that interoperability, accessing all of the da- potentially accessing all of the data you have, right? You have, you have the ability to bring any of
the data that's in your organization in here. And then the second thing is you don't actually have to ingest into our system, right? Or, or how do I put this? You don't have to move the source data. You can take- Absolutely you can, you can access the source data, just bring what you
need to run as RTO of this. But you don't actually have to copy over the source data to get started, which again, would be a month-long, you know, month-long project, that would require a lot of resources. And so, so this is just something you can plug into your existing system, and it will operate within, within those means.
That's very- Absolutely. And so you brought up one of the, like we talked about before, there are some, some non-technical challenges associated with, one, data preparedness, two, actually building enterprise apps. Another one of those, just like you said when you were a data analyst, when I was a data engineer, we had to forklift data- Yeah right, to very specific environments.
Matter of fact, we've got customers right now that I've been consulting with that their entire data pipeline requires the forklifting of data, mirroring of data to different, siloed, enclaves for post-processing. Yep. Right? Yep. And then there's this whole process that happens.
There's a whole economy built around you having to do that within your environment, and it adds complexity. Mm-hmm. So sticking with some of our Everpure principles, this allows you to, This is a tool to break down some of those silos, right? Yeah. You're accessing data where it lives.
Connecting to that data after it's been identified. You have authority to, ingest that data within a pipeline. You can create a pipeline, manage the ingest of that data within the pipeline, which adds a little bit, of additional level of governance and, AI risk management, associated to which data are feeding, AI systems.
Yes. Super, super important. Super powerful. Having, having been data analysts and data engineers- Mm-hmm we, we clearly see, you know, all the, all the time savings and, and all the, the headaches that, that this would save. I know we have just a couple minutes left for the demo portion.
Austin, is there anything you wanna bring us home with? Yeah. Key takeaways, you know, the app is kind of cool. It helps support, proof of value for customers. But what I really want you to see the value of is not only does Data Stream provide the
ability to have an orchestrated and automated data pipeline, from ingest all the way to serving, but it provides, key features in the form of integrated large language models that are, that are optimized for, search and query, concurrency, et cetera, all within the platform. So Data Stream is actually providing the pipeline and an actual AI
capability out of the box. Yep. Yep. Very cool stuff. Very cool. All right. So we will, switch back here really quickly to, a couple slides to close us out. While we're going through these, please do, if you have any questions, we're saving some time
at the end, so, so make sure to, to put those in the, in the Q&A box. So essentially this is Everpure Data Stream, right? It's a full stack, AI data solution that turns raw data into AI-ready pipelines in a matter of minutes instead of months. Austin, from, from your vantage point, everything you showed us, what do you think is
the most powerful capability or, or the one that shows the most promise for, for really helping the customers that you've worked with? I would say, the simplicity of Data Stream. Right? Because some solutions nowadays with AI trial to try to boil the ocean.
Right? But the challenge that will address most of the concerns that our customers have is something that does what it's supposed to do, very well, right? It does it very well, and it helps them ensure that they're meeting all of their, their data and AI governance mandates.
It's providing tools and capability that ease the burden from their teams that are only getting smaller, right, and having to do more work. Providing this capability that is able to be integrated into their current development pipeline and integrate with majority of the tools that they are currently using, in a very simple, way.
And then it also provides capability within Data Stream that allows them to start, truly understanding what is the what are the impacts of those pipelines on an actual, AI system, right? And what I mean by that is we provide that semantic search capability so you can build the pipeline and see if it works.
We provide large language models. You can build the pipelines and see if you're getting the answers that you're looking for. So that accelerated time to value, all within a unified platform, is what I see as the true value for, our customers. Yeah. Yeah, that simplicity. That's, that's a, a, a great way to put it.
Essentially, from my standpoint, it looked like you're essentially connecting to the data, and then all of these steps that you see in the slide below, ingest, curate, classify, embed, index, retrieve, and generate- Mm-hmm all of that is being covered by Data Stream. So, sometimes I think when you see the demo, when it is simple, it's almost like you,
you don't understand what's happening behind the scenes, right? And there's so much to it here. So, yeah, that's a, that's a great way to put it. It's all about the simplicity. And of course, you know, this, this service is built off of the NVIDIA AI data
platform reference design. They clearly saw a need here for their customers. If you think about everything that NVIDIA does, sort of above the, the data foundation layer, you know, their, their services, there's compute acceleration, there's so many things that NVIDIA provides.
But what we bring to the table is really that, that data foundation, that powers AI factories, but also getting the data ready and refined to actually bring into the factory, which is a, which is a really critical, piece of this as well. All right. So we talked a little bit, Austin, about the simplicity and how important that is.
You know, I think, just to, just to drive that point home, what, what this is going to give you is the ability to, control, who has access to data, who ha- control the ability to govern it, and, have that governance built in at the data layer, and also the ability to, to really, manage costs as you're simplifying that, that data pipeline building step.
It's going to get you, get you to, value from your AI projects faster. That step of actually getting the data ready is, one that probably we aren't talking about enough right now, as we talk about AI, because there's so much excitement about the innovation happening, for large language models and, and all of those things, and, and they're so innovative. But when we bring that into the enterprise,
we're faced with the realities of, of having to actually, get the data ready, and, this really speeds up that, that time to value for those AI, projects. And of course, for all of you familiar with EverPeer, you're going to get all the benefits of this being built on top of the, the EverPeer platform, and as part of that, that unified data foundation.
All right. So with that, I do see a couple questions coming through, but there's still time to, to add questions if you like. I'm gonna skip the line though, Austin. I have one more question that's been in the back of my mind as we've been talking about this that I want to ask, and that is: How does the new EverPeer data intelligence service,
so this is the service that we now have after the acquisition of OneTouch, how does that work with Data Stream, to provide, to provide value to customers? Oh, that's, that's a great, great one you just brought up. So, as we're talking about how are we even identifying which data are relevant to answering the questions, those questions being
things that we present in the form of, of, of value in an application or some type of analytics capability or decision support system, or how are we ensuring, the governance? Well, EverPeer Data Intelligence now, we are providing, our customers the ability, to go out and do deep identification of various types of data across their environment.
Those data can be, identified and tagged and classified based off of their governance criteria, and this works across their entire data estate. Once y- they are able to get this rich, really, like, data catalog, they're able to start doing master data management and applying different controls to that data, really verifying and validating access, to that data, and then building,
ontologies and a, a, a knowledge graph based off of that data- in order to start having responsible AI integration with that data, right? So, and I'll just give you a, a heads-up. I was with a, a state and local government customer. They had, some significant public trust issues based off of, you know, some challenges that
they had in the past, and they were like, "Hey, do you have anything that will help us in this space?" Everpure Data Intelligence, checked all of the boxes. So, it's a very valuable tool that can integrate with DataStream, right? 'Cause i- Everpure Data Intelligence, we've got 60-plus, connectors, so ensuring that we're able to connect to all the data across your environment, process that data, support
it, and then connect into DataStream, and then start to, feed that data into data pipelines that are curated, and purpose-built. So it's sort of Everpure Data Intelligence is really helping, with that identify and discover phase, but not just, you know, one time, all the time, right? All the time. You have this deep understanding of your data,
and not just the metadata, but what's actually in the data. And, that's again going to speed up your time to be able to, deliver on these projects and improve governance, and control as well, right? Absolutely. Yeah, yeah. That's, that's very, very cool. Okay, so, one question from the audience, how does DataStream handle governance and
access control across different source types? Yeah, I would say that, right now that is handled via some inheritance, and it depends on what your specific identity access management tools are that you're using within the environment, whether we inherit, different rules based off of your strategy for identity access management and, permission delegation.
But it's all, interoperable and extensible, across your environment to meet your requirements. Oh, excellent. Excellent. Well, I know we are at time, so with that, I'm going to say thank you so much, Austin, for joining us today.
It was an absolute delight. I would love to have you back on. I'm sure our audience feels the same as well. All right. Thank you, Erin. It was a pleasure. Awesome. Thank you, Austin.
And to our wonderful audience, thank you so much for joining today. We really hoped that you learned something, and you got some value, out of the time. If you're interested in working with Austin on a proof of concept, or one of his fellow colleagues, you can reach out to your account manager at Everpure, o- or go on everpuredata.com, and, let us know what, what you're working with, and we can get you in
contact with, this, incredible team of field solution architects, and, and get you going, on your project. And with that, we will go ahead and close for today. If you are interested in continuing the conversation, join us on the Everpure customer community.
Austin is active on it. Others of our, of our FSAs are active on it, so you can have, real discussions, on the details, behind what you're, you're trying to accomplish and what you're trying to do, there. And with that, thank you so much for joining us today.
Erin Stevens
Senior AI Product Marketing Manager, Everpure
Austin Boone
AI Solutions Architect, Everpure
Most teams racing to operationalize AI are stuck on the unglamorous part: getting data ready. Fragmented pipelines, data silos, and brittle hand-offs between data engineers, data scientists, and MLOps teams stall projects long before they ever reach production.
In this session, Everpure experts will demo the new Everpure Data Stream service—our purpose-built answer to NVIDIA’s AI Data Platform and the “data prep problem”—by building an AI data pipeline in real time. You’ll see exactly how Everpure Data Stream helps AI data engineers automate AI data pipelines without manual work.
Key takeaways:
- How Data Stream accelerates the path from raw data to a production AI data pipeline
- Deploying an AI data pipeline on a single, unified platform built according to the NVIDIA AI Data Platform reference design
- Where data-readiness bottlenecks actually hide across MLOps and AI data pipeline workflows—and how to eliminate them
- Why a unified platform beats stitched-together point solutions for reducing complexity, cyber risk, and cross-team friction
We Also Recommend...
Personalize for Me
My Updates
No updates at this time.