Shailesh Manjrekar, Fabrix.AI | NetworkANGLE

[music] Enterprise IT environments are becoming more distributed, more dynamic, and increasingly agent-driven. Yet, most operations teams still work across disconnected tools, static dashboards, and manual workflows. And really, that creates a a fundamental mismatch. Application and AI teams can move faster, but the people responsible for reliability, security, and performance are still spending too much time gathering context and trying to coordinate action. Welcome to this network angle. I'm Bob La Liberte, principal analyst at the Cube Research. Joining me today is Challes Magikar, the chief marketing officer and head of AI strategy at Fabrics AI. and we're going to discuss Fabric's AI's vision for a multi- vendor governed Vibe Ops operational intelligence platform and what that means, why it matters for it and network operations and what organizations need in order to trust agents in production. And Chalesh will also show us the platform in action. Chalesh, welcome. I know you're an alumnest here, but great to have you back on. >> Hey, thank you, Bob. It's been a pleasure to be on your show again. Uh yeah, looking forward to this conversation. >> Absolutely. All right. So, so let's get started. I I gave quite a mouthful there, right? Multi- vendor governed vibe ops operational intelligence platform. So that covers a lot of ground. So I think the first thing we need to do is let's begin by making it tangible. Let's talk about what is governed vibe ops, who's going to be the user and what problem are you trying to solve for for the operations teams? Yeah, that's that that's certainly a long kind of entitle, right? So, let's kind of decipher that what we mean by that, right? So, if you know wipe coding has been a term coined by Andre Carpage, right? And what uh it brought to the light is the ability to actually leverage LLM or the AI coding assistance for writing code, right? And that has become by far one of the most widely used uh and um democratized use case for AI and LLMs. Right now WOPS is kind of an subset of W coding where you're using the same principles but primarily focus on generating operational artifacts. Right? And what what I mean by that it could be dashboards, it could be AI agents, it could be workflow automations, right? Or automation workflows, right? All of that using the same principles and the same tools what you have been used to in the W coding arena. What I mean by that is you could use cloud code, you could use cursor, you could use codeex, right? Uh what have you and I'll talk about we pretty much support all of those leading platforms. Okay. So that's what we mean by uh wops essentially and uh the use cases where wbops at least to begin with is is is kind of envisaged uh to play a role is primarily around the same use cases where we've been playing which is observability which is assurance uh which is sec ops which is change management and so on and so forth right so that's kind of the second deciphering let's talk about now what do we mean by govern so there are two aspects to govern so first and foremost Most the govern comes from the agent tops ability of an harness or of an platform and what I mean by that is the ability to collect to disparate data sources right whether they are APM tool ITM tools npm tools right or just getting telemetry from devices you got to be able to coales all of that that's the first step right so you it has to be your wbops has to be grounded in an customer environment and in in a customer data sets Right? So that's kind of the first aspect and then obviously you got to uh you know make sure that the context is very curated and so on and so forth. Okay. But that's what our harness provides right or what or any other harness is expected to provide to to the operational aspects of wops. The second aspect of governance as I was telling you about was you know w coding by itself has a little bit of bad connotation because the AI coding assistants are generating so much AI so much of code nobody's inspecting that and everybody's focused on the outcome right and that has kind of created created a bad reputation for w coding to some extent with tools like lovable and so on where you know uh there were some vulnerability uh security vulnerabilities which were exposed in this code. So what we are doing with the platform is many of these platforms are integrated with our fabric CI uh agentic platform. There are two in particular which are actually embedded in the platform itself. One is an open-source tool called open code. The other one is IBM Bob because of our partnership with IBM. uh but the rest are also integrated and what that does is now this AI generated code lands into what we call a land landing zone and you are able to look at that code you're able to review that you are able to version control it you're able to test it so we have tools like playright uh which are already integrated uh into the platform so you are able to test APIs and so on and so forth and then eventually you are also able to generate AI summary which I'll show you uh in in the in the demo right but uh this is What we mean by the second level of governance where now we are making this AI generated code enterprise ready. Okay. Uh and and those artifacts as a as a nature as a byproduct of it make them enterprise ready. >> Got it. Yeah. And that makes sense because these aren't people who are actual coders that are going to be using it. Right. This is the operations teams. So providing those additional guard rails are going to give them the ability to trust what they're doing and how they're able to create those new dashboards, those that new visibility and so forth that they need to run their businesses better. >> Yeah, absolutely. So the personas whom we've envisaged using this are primarily you know the operational personas. It could be SRRES, it could be IT operational guys, it could be platform engineers, right? And uh the the the beauty of a w coding is essentially the ability to ex express an intent in natural language right so it doesn't need a whole lot of skills for them to understand it right the use cases remain the same the outcomes remain the same right it's really how you kind of achieve those outcomes right is what really u kind of makes the difference >> okay and then one other thing I want to follow up on you had talked about vaude coding and its in its essence of requiring an LLM to be able to use it and go back and forth So with your product are you actually going out to the public LLMs or have you developed your own internally? >> Yeah. So that's a very interesting question. So one of the other uh um capabilities we are announcing along with WOPS is this ability to have our own SLMs right small language models. So uh obviously there are other players also doing this but what those SLMs provide you is domain specific uh knowledge right. So now the the the AI agents we develop or even the AI assistants who will be doing the wops work they know exactly what is a semantic to talk to our platform. So that's one aspect. So we have an LLM primarily focused on uh enabling the semantic and how to talk to our platform. There's another SLM what we're using is for AI ops. So this SLM understands the customer environment. It understands the customer deployment or its service maps. it understands the correlation policies and so on and so forth. So that's primarily think of that as an as an subject matter expert for AI ops use cases and the third LLM we are we will be launching is primarily around what we call vulnerability exposure right so that's more like an SEC ops use case so as you can see this SLMs bring a multitude of uh benefits to the end customer so first and foremost is what we call sovereign AI that means uh these are SLMs which stay within the customer environment they learn about the customer environment and if you tomorrow decide you don't want them all you do is you just delete the adapter right so these are based on wellproven methodologies like Qur fine-tuning and so on and so forth right the second benefit you you get is obviously the performance because now we are not going out on the wire right over the network and bringing it from an cloud LLM the third uh this is around efficiency right so our accuracy so the accuracy with this SLMs can be extremely high almost up to the 90th 90th percentile because they know your domain specific data. And finally, last but not least is the economics what they build bring, right? So you're not paying up, you know, a whole boatload of money, right, for for the cloud LLM, right? You could use still use the foundational models where they're relevant. So we have an internal router which leverages the right LLM or SLM based on the task at hand. But that's essentially uh the the other announcement we'll be making. >> All right. No, that sounds good. And I think that's going to be important, especially those SLMs that are focused on their particular environments and so forth, as you mentioned, are going to help drive up the effectiveness and the the uh accuracy of those models as well. Um, all right, I want to change gears a little bit. I mean, a lot of people talking about the promise of agentic operations and so forth and how it depends on having agents um that are able to provide accurate context and the ability to work across multiple domains and so forth. How is Fabric's AI connecting to multiple vendor platforms and data sources? Um, right. Building out, I think we've talked about this before, building out a living ontology and then be able to orchestrate specialized agents without requiring customers to centralize or replace all of their existing tools. >> Yeah, I think that that's a very good question again, right? So, uh, as I said, so WBOPS and agent tops has to work in tandem, right? uh agent ops is all about how you operationalize the agents and wops is essentially the ability to provide uh instructions or talk to that agent optops platform right um so uh first and foremost you need to have all your data sources right coales right when I mean a coalist is you don't need to bring in all that data so the way we we do it is uh our data fabric right which has been our claim to fame even when we were an AI ops company right? Allows you to connect to disparate data sources, right? I mean it could be APM tools, it could be ITM tools, it could be npm tools, it could be ITSM tools, right? Uh it could be direct devices, right? So we have the ability if you don't have an existing environment, right, which typically doesn't happen, right? Typically large enterprise customers at least have anywhere from five to 20 tools already in their platform which are domain specific, right? uh but if you don't have we have the ability to go directly uh to the devices and collect telemetry and enrich that right so that's kind of the first step the data fabric does uh you know it coaleses all of this right and then uh the way we do that is also very interesting uh you know when we create this ontology layer which is essentially we call it the GPS for AI agents right so it's a living ontology and the way we do it is through AI agents again right so we don't do there's no hardcore uh you know rules or anything of that kind. We go out to these data sources, we do data discovery, understand the schema of each of these tools and then we bring in that metadata into an interconnected layer. Right? So now the the the agents have some structure when they're looking at at the data level right. So and the way we do this uh you know so it's all agents using MCP and we call this as universal uh universal tooling that means we have the ability to create MCP wrappers run in runtime right whether you have an MCP server whether you have an API or you have just a device right so we can get data create this ontology layer so that's kind of the one of the major building block as part of the harness and then of course the the other aspect is you how we do tokenization right so the context intelligent context engine what I we have has lot of optimizations on how we preserve this tokens right so I I keep saying that hey not all harnesses are built equal right so our ability to give very precise and curated information to this agent provide inter tool connection right provide this LLM router right providing compaction um you know caching at at an context layer is all of those things are kind of you know paramount when you're looking at you know optimizing your tokens. Okay. >> Okay. >> But I hopefully I I address that right? >> Yeah. No, I think that's I think that was good. I mean I mean it's clear that this living ontology layer is very different from a static CMDB or a service map. I mean is it is it fair to think of it as kind of an operational digital twin? [snorts] >> Yeah. So I think digital twin is kind of an uh slightly overloaded term and it's been used in multiple uh you know context by multiple vendors. So digital twin kind of goes a little further right where you are actually simulating some environments with the data at hand. you have the ability to do you know like a DVR move forward move backward in time the ontology layer is primarily targeted for solving one and one problem right that hey providing that accurate and curated information to this AI agent so they are not hallucinating they are optimizing their tokens right so if you think of it you know now say an agent or an LLM is given a task to do a root cause analysis right and there are five or six different artifacts which It needs right say maybe it needs lock from Splunk, it needs metrics from Diana trace right it needs events from some other event uh um kind of an data source right so the ontology layer tells the agents right out of the bat that hey this particular data source is ready it's enriched and it's ready to be consumed right so that's kind of an green indication for the LLM it could be orange where the data source is available but it's not enriched right or it could be white right or gray where the data source is absolutely not available and that's where the agent will have to do some MCP calls and and so on and so forth. So this is how we kind of provide that intelligence or that semantic layer or also the context graph if you will to those those agents. Okay. So slightly different than uh digital twin itself and it's kind of an overloaded term in my mind. >> Got it. No understood. That's that's some great detail on the on the product itself and and clearly, you know, having the right architecture is important, but buyers also really care about their outcomes and the operational outcomes. Can you walk us through how this model changes real a real workflow? You know, maybe it's uh network health VPN troubleshooting, CVE exposure, something like that. maybe a full stack root cause analysis as it compares to maybe some of the traditional swivel chair operations that are occurring today. >> Yeah. No, I think that's that's really good question, Bob. Right. And you know, it finally boils down to uh what's your success story, right? And we are actually we we uh we are very uh you know um uh kind of uh we take pride in actually sharing that we already have couple customers who are leveraging the platform and have seen some tangible results. Right? So there's this one large fortune 500 customer which is uh using us in production almost from the beginning of this year and what they had before us right so think of it as an anc which is based on you know the traditional tools right they had about 25 different disparate tools right right uh they had about 12,000 or 13,000 assets about 207 applications right and they had this mandate to go agentic but they were stuck with this what we're calling in the swivel chair problem, right? Where you have different dashboards from different vendors and you are in a war room trying to, you know, with five different subject matter experts looking at those disparate dashboards trying to arrive at a root cause, right? We were able to eliminate all of that with this W operated dashboards, right? which pro provides now a single pane of glass across their you know disparate environments across applications across the agent technic across infra across service ops right and we did an analysis of that right uh uh without agents and with agents right and particularly agents which were built uh with uh wops right with the dashboards and so on and they were like they had phenomenal success right so they were able to almost save about uh 1.23 23 million per year uh with this approach across full full stack root cause analysis right VPN uh right Wi-Fi and then also CDs right and what I mean by that is the previous uh workflows were you know you're looking at all this 20 25 different tools right you're spending time you know getting those artifacts you're kind of trying to converge on a root cause analysis now you have agents doing all of that right when it comes to VPN and and Wi-Fi right disparate different tools right you have uh Cisco you know u the the firewall right asd right uh you have uh you know identity management systems across all of this what we had was what we call ambient agents that means these are proactive agents which are continuously running and monitoring and Wi-Fi point endpoint uh it's w it's monitoring uh the VPN environment right so this particular solution we call it as deex X digital uh employee experience right so it's monitoring all the way from the endpoint to the data center right across SD vans and and so on across almost eight different domains and uh we were able to create and help desk without an ticket for them right so it's a very powerful statement right as you can imagine that there were no I mean because those proactive agents were looking at um at those issues you know before they could even arise and they were able to some of them they were able to fix and then you know uh either have human in the loop to resolve to final remediation step or at least notify them right so those were kind of the outcomes where which we were able to kind of prove the second customer is actually a large SI right and they are also seeing uh they started with their internal IT operations right and they were so so happy with the outcomes now they're training their full kind of strength of fds what we call uh you know full stack full dep uh fullstack deployment engineers right um where which are kind of expected to be the front end to take an AI use case and and provide an and deployment right so they're training those so those are some of the outcomes which I can talk to kind of share with you Bob >> no absolutely those those sound great and actually my next question kind of builds on that I mean you're giving operations teams the ability to create new dashboards agents right and a lot of automations right really powerful stuff, but many operators today are not developers. So I'm wondering, especially with those customers you had mentioned, what does that adoption journey look like and how do you help teams become more comfortable using these capabilities safely in production? >> Yeah. So that's that's a really good question, right? So if you look uh you know with with the wipe wipe ops techniques, right? So there is really not whole lot of uh you know expertise needed to actually create those artifacts right now. How to use that artifacts? You need subject matter expertise right or actually telling that intent. So that's where the subject matter expertise comes in right and that's no different than what they had before. So you know the the the the personas whom we tend to go after platform engineers or S surres all right and they are now increasingly not just you know managing operational aspects but also uh responsible for you know devops and so on and so forth right so it it kind kind of comes uh natural but uh to your point yes uh there there would still be subject matter expertise needed on how to use it and what what outcomes uh you have to uh look for and kind of work towards Right? So that's one aspect. The other aspect is is clearly you know this FDES right which is becoming an prominent role now in the AI arena. Right. So this FD so we've been starting to train FDs with our SI partners right and the OEM partners alike okay uh who have a better understanding of the AI domain and the AI landscape right and they are able to handhold the customer and we do the same thing. So we are a relatively small team but with the this large customers whom we are talking about we handhold them through their journey right so again this is a new arena but you would see uh you know Gartner has put out almost two publications around wops and how wops is relevant for an agentic knock or agentic sock right so the underlying infrastructure is changing very fast to keep up with this new approaches around wipe coding and agentic and so on and so forth right and that's where I think you you were asking me earlier uh it also equally applies to uh you know um to companies who are not using wipe coding today right I mean they could just be doing agent tops right and this principles are still relevant for them >> got it no that makes sense so certainly you know giving the forward deployed engineers some some tools and some technology that's going to help them enable that given all that when you get in you get started what should a customer expect in the first 30 60 90 days what's the time to value for organizations deploying the platform >> yeah so I would I would suggest you know start with smaller use case right where you can identify you know uh tangible goals what you want to accomplish right I mean it could be a VPN environment right uh it could be your Wi-Fi environment right uh and or it could be uh just your application start with one application monitoring right which is of importance to you, right? Um >> start there, you know, do a pilot and and then as you know, start with productivity enhancements. So that's kind of the first step where we see you know the value in the 30 60 90 days. uh as you kind of evolve from productivity enhancements then you are able to see that hey you are now able to consolidate some of these tools right that hey you don't need 20 25 tools at all different levels right then as you progress you see that hey many of this you know uh the the uh the environments or the traditional disciplines are actually collapsing right because now like you know when you think about reasoning model like mythos or daybreak right these reasoning models they understand boundaries right I mean you you are getting so many day zero vulnerabilities right out of say a mythos now you need to understand using the sec ops principles that hey what is actually what does this mean for me right what is my exposure what is my blast radius uh that kind of leads you into the IT ops arena where you're looking at all your deployed environment right and then it kind of crosses over to the sock right which has been traditionally where you have been looking at vulnerabilities and so on. Okay. So that you are kind of you know collapsing this this kind of you know boundaries where IT ops sec ops right knocks are kind of consolidating under that same data layer or that ontology layer. So that is how we see the progression happening across uh different customers and and these personas. >> Excellent. That sounds great. Well we we've talked about the product and what it is. We've talked about the architecture and the outcomes and as I alluded to in the beginning you know let's make this real for the viewers and chilles can you give us a demo of the platform >> yeah so Bob uh yeah I'll certainly try this is a live demo right but uh yeah I would be happy to share you know some of these capabilities uh what we are bringing to the market all right here hey folks yeah so here's uh here's a demo Bob to kind of illustrate what we mean by u you So I'm going to launch this. This is a real live demo. I'm working on off of my VPN here. Okay. So things may be a little slow. So this is our agentic platform. Okay. Um and you can see this uh this kind of shows you uh you know the dashboard across you know the hierarchy of the agents what I was alluding to earlier. So these are all white coded applications on the top right and then you have this highle agents which then in turn talk to the use case agents right whether it is ITSM as an use case AIO ops as a use case anomaly prediction right the digital sur in turn talk to the tools and platform agents right where a VPN health agent knows how to talk to say the Cisco firewalls right or the Wi-Fi tools right a root cause analysis agent knows how to talk to a Splunk Dina trace right and what have you right and then on at the bottom you see this this dynamic integrations through our uh dynamic MCP server but just to illustrate what I mean by ancoded dashboard right here is an example of network health dashboard okay so this is a dashboard which is interactive right and you can see it is actually showing a disparate different devices right so you can see 354 devices devices downs plug indexes active issues affected side and monitored platforms are Juniper Mist, there is Mari, there's catal catalyst center, there is NDFC, there is thousand eye, Splunk, Cisco FMC, uh there is service now. Okay. Uh there are it's flagging all the critical issues. It is showing the total incidents. It is showing latest team notifications. It is showing the VPN, right? Uh it is showing affected sites and so on and so forth. So this becomes your kind of a single pane of glass, right? So this is what we mean by you don't have to get into that swivel chair operations right so you have a single pane of glass where we are able to get almost all of that information for you and these are interactive right so you can say hey I want to double click on this right say uh and you can get to the next level right so for example here's a catalyst center right you want to double click on this and you can make and change this on the fly it's no longer you know another week and and and a month to develop a dashboard, get it permitted, uh you know, get all the right permissions and so on and so forth, right? So, you're doing this on the fly. The second thing I wanted to show you was how we actually create these dashboards, right? So, uh bear with me, right? So, this is what you know what we call uh WOPS, right? So, these are all the all the dashboards which are created. Okay? And you can preview them, you know, you can share this with somebody else, right? So this is the other governance part what I was telling you about Bob right so now we have made this course uh the AI coding assistance whether it is so right now we support uh cloud code we support codeex we support a cursor right uh we support uh you know um uh Google anti-gravity and there are two which are more closely integrated one is open code and Bob IBM Bob and why did we do that right see because uh you know each of these tools need a license so some of the operational folks came back saying that hey we don't want to spend on those licenses right so for them we are using an open source version which is equally good enough right and embedded in the platform so you don't have to use another interface but here you can see you know so this is an AI summary which which it has generated for this particular dashboard right what was the action what were the changes right what were the streams used you have all the artifacts around this right and then you can look at the preview of this dashboard right so This is a dashboard right or you can open this. Okay. So this is our landing area. So once you have generated that code right the code lands into this landing area. You can inspect the output. You can actually test it to the to to your liking right um and so on and so forth. And you can see there are multitude of this um you know um dashboard. This is AI observability dashboard right. So so this is what we mean by wipe coded dashboards. Okay. Uh hopefully it's it's loading on my VPN. So we'll come back to this right in the meantime. I I'll show you you know how this all all works around. So in my uh this is my cursor environment right um right uh this is my cursor environment and uh the the only prompt what I've given to it is you know uh create a dashboard to plot time series graph of UDP syslo injection right so this is a p stream and it has gone and it has actually worked for almost 2 minutes and 47 seconds it has looked at the dashboards it has looked at all the data sources it has access to okay And um it has actually come back and said hey to deploy the agent use this to push to the landing zone. So only when I execute that command the push is succeeding now right and now this dashboards get created right. So this is this is cursor environment right another uh this I have given it is hey list every pream available to me in my environment. So it it knows how to go and query because you know we have provided it the grounding as to how to talk to our platform and how to get that grounded data right so it is saying hey there are about 463 P streams okay and it is showing me all the all the capabilities around what are those P streams what are the data available and so on and then this is the platform where you can say so this is Bob AI right so this is IBM's Bob it's an again AI coding assistant which is already embedded into the platform. Okay. So you can see here and you can say hey connect to the Bob right and you can give it similarly whatever I was showing you. Uh you can give the same commands to it right and so on. So uh you know so this is actually an and um this I had I had executed before right. So generate a dashboard right. So fabrics via artifacts it it generated that particular dashboard. Okay. Um so this is kind of the power of what we mean by you know uh another thing I wanted to show you if time permits is um you know so this is open code right so again um uh a simple simple kind of an prompt okay so generate a dashboard for pstream fabrics vx artifacts right so this is again built in into the platform okay uh and then it will ask you some questions if it has to right and then you know you can eventually push this into the system so this is how you use different kind of coding assistance right uh and tools. Uh let me see if I have anything more to show. So there is also an data exploratory uh agent what we provide right. So you can explore the data before you want to start doing this right. So it goes and actually looks at the preams right. So you can choose a P stream by name. Okay, say you want to look at uh u CD uh let's look at net network devices interface, right? And it will actually show you that data set before you actually, you know, decide to uh Okay. Well, let's look at another one. [clears throat] So yeah, I mean while it is doing this, so you can see the data, right? And now you can actually set up right. Another thing I wanted to show you was you know wipe coding right. So these are some of the dashboards right. So this is something which I have done myself. So my dashboards these are our development guys are doing this right. So say if you want to look at AWS right EC2 capacity. So it will show you the AWS EC2 capacity right and then you know here itself you can look at hey I want to test this using test cases or I want to review the code right so this is the other layer of governance where now you are making this AI coding assistance enterprise ready right you're making them you know that's what we mean by governs okay >> um so and then this is the AI summary right so share dashboard you can share it with whoever you want to and so on and so forth. So that was a quick preview. Um Bob, right? Uh but happy to kind of talk more at splunk.com. Um okay, so I'll stop my sharing here. >> Yeah. No, that was that was great. I think it gives people a real idea of just how quickly you can go and create those those dashboards, how you're able to test and validate them out and then actually get real results that will help you operate your environment. So I I think that was great. And I think for me I mean the the broader takeaway is that agentic operations it doesn't simply mean adding another co-pilot on top of another dashboard right to create meaningful operational value right those agents need the trusted context it needs that cross-domain reach there's got to be clear governance and a practical path for humans to remain in in control of this which I think you you've put that together so I mean your approach is designed to help operations team move from manual correlation and reactive troubleshooting into more of a a curated agent-driven workflows across all the tools that they already use. And so I think you know the proof point for enterprise buyers will be how quickly the model can deliver measurable outcomes in production while preserving all the controls required for reliability for security and cost management. >> Yeah. No absolutely absolutely right. Right. So our as as we began this program right so our vision around and successful wops operation are there are three pillars to this right so the first is you got to have you know the right model for the right task obviously for sovereign AI but also for the economics uh the efficiency right and and um essentially the performance and the accuracy of it right so that's kind of the first pillar you either use an SLM which is domain specific or foundational model where they relatively they are applicable. The second thing is a good harness right. So this is where the agent tops capability what we've been talking becomes paramount right. So this is an platform uh which has all the bells and whistles right our own observability explanability we have our own eval modules we can also use other eval platforms like whether it be ARS from Detrace or Galio from Cisco right uh and that harness needs to be able to be open on the on the bot on the top right where we could use any AI coding assistant it needs to be open in the bottom where it it can interface with any data source right so that's kind of the second important requirement and the third important requirement is what we call tokenization right so because as I keep saying not all harnesses are are created equal so the optimization you put in to be able to save the tokens right because you and also the guardress right so you don't want a rogue agent you know consuming all your budget on all your tokens so we have phops built in into the platform so uh so those are the three capabilities or three pillars what I feel are absolutely paramount account for an successful WB ops capability. >> Awesome. Well, I think that's a great way to wrap up the the video with those three points. Shalash, thanks again for for joining us on this network angle. Appreciate you being here. >> Yeah, thank you Bob and see you guys at Splunkcon. Uh as Bob said uh you know in Denver uh in two weeks uh in booth we are in the ISV platform zone and come talk to us and see us. >> Yeah, absolutely. So, I'm Bob La Liberte from the Cube Research. Thanks for watching. And as Shalles mentioned, if you want more information on Fabrics AI or to get a more in-depth demo, stop by their booth. It's uh 011 at the splunk.com uh conference going on in Colorado from September 14th to the 17th. Or please visit their website.