 It's really fascinating because of what we do, folks who are pushing the envelope are coming to us to find ways to not spend all of their time doing this mundane task of shuffling data around. Hi, this is your host, Dr. Bharki, and welcome to a different newsroom. And today we have with us David Flynn, CEO and founder of Hammer Space. David, it's great to have you on the show. I'm glad to be here. Thanks for having me on. Yeah, first of all, congratulations for 56.7 million funding around. Before we, of course, talk about this funding and other aspect, you know, it's a good idea to just, I mean, we used to cover, I mean, not we used to be covered Hammer Space on a regular basis. But since you are here on the show and you also founder, so I would like to just talk about when, of course, when Hammer Space was created. And now when we look at now cloud native word has evolved, Kubernetes native word has evolved. So let's talk about Hammer Space, the evolution when you created and now where we are. So just talk about the journey quickly. I had the fortune of being sort of front row seat with the introduction of solid state in the data center, which of course, adding a whole new performance tier drove the issues of data being a highly localized thing that as a problem became keenly aware when you want to spread data between disk and flash and when you want to spread data out across the servers with flash in the servers themselves where you can get the highest performance. So, you know, we set out on this journey a long time ago with an understanding that bringing a physically distributed data into a logical centralized single file system was going to be important. But little did we know, you know, the trends of cloud computing, the need of hybrid cloud, multi cloud, even just multiple regions within the cloud or even multiple data center, you know, those things were getting hot at the time. But I'll tell you what has really changed recently is with COVID, the need to have a highly decentralized organization so that you can support people, not just working from home down the street from the office, but working from their home country, right, on the other side of the world, and having teams of people in different cities around the world, you really have to tap into the global talent market. And if this is data intensive work, they need to be able to interact with it with low latency and nearby. So that's number one. The other thing is that we have seen major shifts in supply chains and difficulties in getting gear. At the same time, you need ever more exotic types of processing. Now GPUs are very big. It's not just CPUs and these specialized hardware, hard to come by, right, with chip shortages, and you definitely need to move to a cloud model. Let's call it a least model, a time share model with these, because the other thing is another backdrop to this is AI and other workloads that are bringing high performance to the forefront. Used to be high performance was a niche thing, and now it's the mainstay. And so if you look at all of these trends, whether it's the work from home or the need to use third party infrastructure, the data intensive bursty nature of the HPC workloads, all of these mean we need a new way to work with data. The old model of store and copy to preserve and access data just doesn't work. We have to move to an orchestrated model. What I also want to understand, as you rightly said, we kind of live in a data driven work, to be honest. And we say data is the new world. Yes, it is, but all capacity is limited either data is the new solar. I think it's renewable. It's going to remain even after we are gone. But we are not only consuming messy or more data, we are also creating messy or more data. And I'm not talking about consumer space. I'm talking about enterprises space as well, all the way from, of course, Tesla's, even to the ring bells that people's faces are, you know, IOT. We're also talking about we used to talk about big data. Now we are also talking about small, wide data as well. How do you have seen the evolution of data itself? Well, I think what we're seeing is a move to where all data is useful. In the past, you had highly structured data databases, relational databases. Then we moved to NoSQL, you know, now with AI, ML and modern analytics, your graph databases. And basically, all information is useful because technology has gotten advanced enough to be able to make logical sense of these things, to be able to perceive images, those types of things. So now the bulk of data, because remember, only the minority, maybe 10% of data is structured. Everything else is unstructured. So I think one of the main shifts that we're seeing is away from structured data as the focal point to where now it's all data, including unstructured, maybe more so on the unstructured side. So that's one major shift. Another shift is around the lifecycle of data. It's not just interesting when it's recent. It can have a resurgence in value sometime in the future. So it's not just about active archive. It's about keeping everything online and consumable because it can provide a foundation to interpret other things. So data doesn't just have a linear lifecycle where it goes from hot and useful to cooling off and cold into the long tail. So in other words, we're talking about the breadth of data now, including the majority, the unstructured data. And we're talking about the depth of data across time. Those are some major shifts that we're seeing. I think very relevant to the discussion. And another thing with the data is that, of course, cloud native Kubernetes itself is intimidating, but data is not that everybody wants to touch. Also, there is a kind of a scarcity of data scientists, data engineers. It should not become a barrier for entry for even new organizations when they do want to. So how do you see Hammer Space or other players in the space kind of lowering the barrier of entry so folks can start data operations without having to worry about all the mumbo jumbo and plumbing? In two ways. Number one, folks who are in the data janitorial business in the old world where data was made permanent by storing it and made usable by copying it to different places. And that's what most people know that they don't know another way. But we are moving to a new era where data is orchestrated, just like your consumer data on your iPhone or your Android phone. When you go from one phone to the next, when you go between your phone and your laptop or your tablet, you have this expectation that your data is just there. It's orchestrated on your behalf for you. That's what we do with the reams of unstructured structured data, all forms of enterprise data that now becomes orchestrated across different points of infrastructure at scale, directly consumable. So moving from a world where data is made permanent and useful through store and copy to a world where data is made permanent and useful by being orchestrated is a very different thing. And what that allows is for the first time, the very manual process of choosing where to put the data and when to pick it up and move it. That can now be done using automation and in particular AI and ML. It's interesting that we're talking about using artificial intelligence tools to make decisions about where the data is placed, when it's moved and how it's made available. And this just helps tighten the feedback loop using AI and ML to solve one of the longest standing, most challenging problems of our era. I mean, there's nothing more digital than data. And yet there couldn't be anything more manual in the IT world than the choice of where to put it and the downtime and the migration. And when you move it around, it's so disruptive. So I'd say two things. Number one, this allows folks who are coming into the data science business to have to not have to spend their time worrying about getting the data into the systems that they want to use. So they can focus 100 percent on data. The getting value from the data, not getting the data to the machinery that's going to do that. But this is also about taking folks who maybe used to be more on the infrastructure side, having to manage and do the data janitorial and it frees them up to shift. So it's like the human talent can now move away from this other stuff and go towards the value extraction instead of the plumbing and the janitorial. Last thing we want to be stuck in the janitorial work of moving data around in the plumbing. Right. And that just frees up a bunch of people to do something more advanced. If you look at the AI. One is the workloads that folks are running, leveraging these. And second is how the AI can actually help hammer space. So can you talk about these two things? What does it look like for hammer space? So the workloads generally involve three phases. One is ingest. And that's to get the massive amounts of data into the GPU array, typically GPUs. The second, and that's done in massive parallel, a lot of redundant data. You're sending the same data to lots of different GPUs. This the next thing is checkpointing. When you are doing a training job that can cost millions of dollars and run over a long period of time, the last thing you need is for one of your components to fail and to cause you to have to throw away all the work. So you periodically checkpoint from which if you have a failure, you can roll back and restart with substituting gear that's not broken. That allows you to not have to throw away the whole job. But those checkpoints are very bursty and the entire system has to come to a halt while it dumps everything. So it's like, let's dump the state of the world at very bursty and it's a right workload. And then, of course, you have the once the results are done, those are generally more concise, but the output of the training has to be stored. And then so if you can think of those as the linear thing from the start to while you're running to when you're done. But then you have the iteration of your data scientists who need the scratch pad to work with potentially petabyte scale data and interact with it. And there is still no better abstraction, no better interface than to treat it as a file system with direct restructure and so forth. So those are really kind of the four things, getting the data in, capturing the checkpoints, the data out, and then the iteration and refinement, filtering, modeling and running a whole plethora of different tools to filter those data sets. So that's kind of the workload side of it to answer that question. If we answer your second question around how does this help hammer space? Well, it's really twofold. Number one, hammer space is unique in that we pull the file system out of the storage infrastructure. So we are the first company that have built a file system that is capable of spanning infrastructure of any type from any vendor, file blocker object from any of the vendors of block storage, file storage, object storage, whether it's on-prem or in the cloud, literally any kind of infrastructure from any vendor. For us, those are bookshelves shelving on which your catalogs of your reams of data can sit, but the actual logical structure of the data, the file system, the directories, those are the same regardless of where the data is at any given point in time distributed across that infrastructure. So what this means is that for the first time, you can automate the granular movement of data. And we do that in both a push and a pull model. Through policy, you can have data pushed to the places where it's needed in advance of actually needing it, to the cloud from one cloud to another from one data center to another, et cetera. And you can use AI to help identify those things and get them moved. And the second is a reactive model, a pull model, where the very act of trying to access it will do the work to go and pull it. And because it's very granular down to the individual files, you can work with massive data sets. Now what this means is that you never have to stop your use of the data to be able to move it. We're actually assuming that the base state is that data is always in motion, that data sets are going from one place to another and around. And that's what we mean by data orchestration. And it's fundamentally different than data management because management assumes you have to stop using it to copy it. Then you can start using it. It's very mutually exclusive. The movement of data is exclusive to the use of data. This solves that. So in our case, AI can be at the heart of what I was talking about, the data janitorial work, the movement of data, the scheduling of how things get moved from one site to another can finally be automated and free up folks from having to work down at the infrastructure layer to now where they can work up at the platform and the data layer. And since we are at one point talking about new workloads, of course, Trinity AI is there next year when Apple comes out with their Vision Pro, it will open a totally new market of VR and once again, VR, everything is cloud based data. Of course, we talk about EVs, cars are there. What other industrial, once again, new kind of businesses that you are seeing, which are data driven, once again, which you see as opportunity, as well as challenges that, hey, we will have to solve these problems. You know, it's really fascinating because of what we do. Folks who are pushing the envelope are coming to us to to find ways to not spend all of their time doing this mundane task of shuffling data around. And one of the things we're seeing, so we get to see all sorts of interesting business models. And one of them, and I forget the name of the company, but they're putting in this, it's a stadium. I think one of them is going in in Las Vegas. It's a huge dome and it's video all the way around completely immersive for crowds of tens of thousands of people. You know, to do event types thing in full surround. And that has with it massive amounts of very high resolution visual data. You know, that's that's one of the things that we see. Other things we see are different business models around. I mean, things is as as simple as doing high performance graphics workstations as a service in the cloud. But with the data being orchestrated so that your on-prem file systems can also be present in the cloud and in different regions or business models where you're now able to hire talent in, you know, remote regions of the world and have cloud or even colo facilities nearby that now have a presentation of all of the same data with high performance. So now if you're editing movies or designing microchips or designing spacecraft or autonomous vehicles, all of those things now you can you can tap into the global talent pool and have them readily productive because they have access to the data nearby wherever they are. One more thing that I want to talk to you about is that when we talk about data or when we talk about this whole space, of course, you know, ever since our space came into existence, we also talk a lot about cultural shift, cultural, you know, movement that are happening. We mostly the most popular one is the whole DevOps movement, but we do talk about, you know, SREs, shift like movement. How important is it for organizations to also have a culture which is once again around data because data yes, of course, in traditional terms, it's a silo, but today it is starting the developers. That's where they start dealing with the data and it goes all the way to the ops and security security folks. They are more concerned about saving data applications can come and go. So how much culture shift you're seeing? Or you feel that no, we have to do more work to build data centric culture within organizations. Absolutely. You know, there is a mindset. It's human nature to live inside the box that we are aware of. It's the Plato's cave problem that we see the world, you know, in the two dimensional projections on the cave wall, don't see the world in 3D. And in the world of data, that is evident in the fact that people assume the boundaries that data is a highly siloed and highly localized thing. And when you introduce the new degrees of freedom of having data readily consumable at high performance across multiple systems, even in different data centers, that is mind bending. It's like introducing a third dimension where you only had two. And it goes to, by the way, some of the same principles behind, and I mentioned this because you mentioned it, Kubernetes and container orchestration and microservices. The microservices container orchestration paradigm is a mind, a cultural shift away from monolithic virtual machines and virtual machines were a culture shift away from, you know, physical machines and feeding floppies or feeding CD ROMs into servers that you had to rack and stack. We shifted to where now you can spin up a server virtually and it changed the whole paradigm for deploying servers. And now with containerized microservices and orchestration, it's changing that yet again. Now you've got principles like granularity, lightweight encapsulation, statelessness, and that allows you much more flexibility at building, you know, systems that scale out. They're not these monolithic, you know, VMs, right? And so we're seeing sort of the second phase of the abstracting of compute away from the computer running it. The first thing was to pick up the OS as a whole. And now the next is to take the actual service and wrap it in a lightweight container and orchestrate it. So it's a progression. And we're seeing the same thing from the data perspective, but people are still used to data being in the old world, you know, very physically bound in the infrastructure. And so part of our main burden challenge is to get people to imagine a world where data is not an emergent property from the storage. It's not something that's rendered by the storage, but data is something that transcends the infrastructure, which is kind of ironic, because data is a higher level abstraction, data is a platform layer thing, whereas storage is an infrastructure layer thing. But because storage has been the owner of the file system, it has owned, subjugated the data, you had data being subjugated to storage, therefore, platform is being inverted and subject to infrastructure. And that fact that you have a platform layer thing that's bound to infrastructure is what has made it really impossible for us to finally move to a world where infrastructure is interchangeable, and where you can run anything anywhere without having regard. And you can see this just by the fact that you have to run your stuff in close proximity to wherever you stored it. That whole principle of move the compute to the data is actually couldn't be more wrong. Data is the thing that's digital. It's the thing that ought to be able to be moved. And yet, the the conventional wisdom has been moved the compute to the data. Well, the compute is the thing that requires more and more exotic processors like GPUs. And it's what you have to actually plug into the power grid. It's what you have to have physically stacked. It requires energy, something from the physical world to actually run it. And so data is what ought to move to it, not the other way around. And yet we are so stuck in that mindset that data is so hard to move, it's so disruptive to move it, that we have to move the workload to the data. And that's just an artifact of having platform being bound and inverted under infrastructure. Now I will talk about the the investment round of 56.7 million. Talk, first of all, if I'm not wrong, in May, you folks acquired a company as well. So I also want to talk about the evolution of hammer space as well, these acquisition and these fundings. When I left Fusion IO about a decade ago, it was with an eye single to solving this issue of data being a prisoner to infrastructure. And, you know, slowly putting together the building blocks and driving the industry standards like parallel NFS and the open source implementation and building and collecting these pieces. And, you know, fortunately, my success with Fusion IO and my background in systems and software allowed me to the resources and the ability to attract a team that was capable of addressing this. But building file systems and then putting the data movement into and behind the file system is just incredibly challenging. File systems have been on a march from being localized inside the OS like DOS was named after a disk operating system. The fat 16 fat 32 file system was the defining feature of that OS so much so that they named it the disk operating system, you know, but the file systems at the heart of Windows within TFS and in Linux, the the ex T's and, you know, file systems started as part of the OS and then we put file systems on the network, like with NetApp. And then we made file systems super scalable by dumbing them down with object storage as three super simple storage, right? As a way to get to cloud. But file systems have been on this march to being ever broader and more decoupled. We finally take that to the extreme where you have a file system that can sit atop anything. And that that was a major amount of work that has required massive investments. And, you know, luckily, my success has allowed me to do that. So, you know, for example, the funding until this round, I personally financed the company to the tune of nearly $20 million. This was not your your standard startup because it does take so much work to build file systems. And we're talking here about the file system and the data platform that, you know, becomes the backplane to a decentralized cloud to a super cloud. And that that so this this announcement now is very important because it represents the first time we've taken on institutional investors. But it's more like a C round than an A round because I and angels that had worked with me through Fusion IO, for example, and did very well, they backed us through the A and the B. And by the way, it wasn't without without challenges. I mean, that's, it's tough to be, you know, spending so much time and money on the engineering side. But the beauty is, we now have the proof points. I mean, when a company like Blue Origin, Jeff Bezos own company, is using Hammer Space as their data platform for their on prem infrastructure and the ability to move it into cloud. So their hybrid cloud solution, that is a very big endorsement. And, you know, that's very exciting for me to have the proof points now that you can actually do what was the thought to be impossible because of the massive amount of investment that it would take to build it. Since we're talking about the investment, can you talk about, you know, what type of investors are there who are, you know, investing in a company? We're privileged enough to be able to wait before taking on external partners. And that allowed us to be selective. And one of the key selection criteria is that we had along the way is that these be long term investors. Now, what I mean by that is investors who would hold and grow with the company even once we are public and liquidity is available, right? So the thing about venture is that they get measured by the returns that they offer their investors, the limited partners. And so there's often a hurry to liquidate and to hand either the money back or the shares over to the limited partners. And that can lead to, you know, folks who don't understand the long term vision of the company and a tendency to liquidate and sell because it's no longer being managed, you know, for you. And it's because venture has a specific mandate and that's up to, generally up to a point of liquidity being available. But what we were looking for is companies that are here to create value in the company over the long term, even once we are public, because for us getting to an IPO, getting to where we are a public company is really the starting gate for the kind of growth we need to see. This is about redefining how data is made permanent and how data is made available for use away from an infrastructure centric model. And that's going to take a major investment in time. And the last thing we want to have to do is to worry about investor turnover. So the thing about prosperity seven, but especially the likes, well, prosperity seven is has one LP, one limited partner, and that's Saudi Aramco. And they are investing for the long run, they're playing the long game because they're investing money in in future technologies. And then arc investments, that's Kathy Wood, you know, she has created a whole new asset class, the disruptive innovator asset class. I mean, very few folks can actually say that they created a new asset class of investment. And she really has it's not about small caps, not about large cap. It's about companies who stand to disrupt entire industries. So it's a very big privilege to have her jump in and she jumped in personally early. And now with their crossover fund, she has the institutional money from arc that has the benefit of the mandate is to is mostly in the public markets to hold the equity. So they just now have a vehicle where they can jump in earlier. And that's the case with all of the investors that we've taken so far is that their main investments are in the public markets. David, thank you so much for taking time out today. And of course, walks to the journey and also evolution of data, of course, talk about the investors. It also gives a very good glimpse that when we look at hammer space, it's not, you know, you did not create a company that becomes a target for acquisition, you are playing a long term goal, which means, you know, a lot of things are in the pipeline that will be talking about. Thanks for all those insights. And I would love to have you back on the show. Thank you. It was my pleasure. Love to be back.