Episode 93: Robin Purohit on Bringing AI to the Enterprise

Brian Dawson: Hello. Thank you for joining us for another episode of DevOps Radio. I’m Brian Dawson, and today, I’m joined by Robin Purohit, and he is the CEO and Co-founder of Peritus.ai. Hello, Robin.

Robin Purohit: Hi, Brian. Thanks for having me today.

Brian Dawson: Thank you for joining us. So, I’m happy to have Robin here with us. As usual, I’m gonna be, based on his breadth of experience and all the questions I want to ask him, I’m gonna be challenged to squeeze all of this into one episode, but we will do our best.

I don't think I could do justice to introduce, really, the vastness of your background. What I will say is, Robin, you have held leadership roles or had a role at a list of who’s who in the past 20 years of tech, frankly—Hewlett-Packard, Mercury Interactive, VERITAS, BMC—big hitters. And then now, today, you founded a company, one of your multiple startups that you founded, that is moving into the important and hot space of today, artificial intelligence.

Maybe you can tell us a bit about what you’re doing today with artificial intelligence, but moreover, as you’ve taken that rich journey, how did it bring you to where you are today?

Robin Purohit: Yeah, that's awesome. So, in my current role as CEO at Peritus, I’ve kind of really been interested and curious about how to bring AI to the enterprise. And so, I advise a lot of companies and startups that are trying to do that. I kind of fell in love with the idea of bringing Peritus to the world of tech support. So, how do we bring machine learning and AI to help engineers do their jobs better?

Robin Purohit: And this probably goes back to just starting my career as a hardcore engineer and doing everything from, you know, chips and high performance communication systems and complex systems software. You know, the world of technology has just accelerated so much, that it’s really hard, I think, for developers and engineers to stay on top of things and be effective, especially as there’s an insatiable demand for more of them in the industry.

Brian Dawson: Right.

Robin Purohit: So, we think there’s a very rich market opportunity to build a family of AI systems to help engineers be more productive.

Brian Dawson: Awesome, awesome. And so, when you started, as we spoke about and you referenced, you started your career as, I’ll quote, a hard core engineer.

Robin Purohit: Yeah.

Brian Dawson: Can you share a bit about what brought you into computer science and what you were working on when you first got into the industry?

Robin Purohit: Yeah, I kinda tripped into computer science. I’m a natural kinda curious person, so I’m always trying to learn new things. And when I was in high school was when the first desktop computer showed up, and nobody knew what to do with them. Everybody was buying them and nobody knew what to do with them. [Laughter] And quite frankly, there wasn’t a lot of software around back in those to make them do anything useful.

So, I kinda figured out how to be a programmer pretty quickly, and then I started getting all sorts of gigs to teach people how to use computers, buy computers, and then build apps for them, whether it was for the home or for business. So, I just became a programmer and consultant out of the blue when I was like, 16, 17.

Brian Dawson: Wow.

Robin Purohit: And I got the bug, you know, I wasn’t really sure what I wanted to do until then. I thought—okay, well, if you’re good at kinda learning the new thing and you like to solve problems, this seems like a good way to spend your time.

Brian Dawson: So, I have to ask, what was the first computer? What was the computer—for me, it was Apple IIs.

Robin Purohit: Yeah, the one that kinda was the hot one in my school was the TRS-80, the Radio Shack computer. And then I talked my dad into buying the IBM PC One, I dove into that.

Brian Dawson: And I know I’m ratholing you, here—what were you coding in when you first decided to write something on that?

Robin Purohit: Yeah, so, I started in BASIC and then we realized pretty quickly that you couldn’t make the computer do anything fun without doing assembly language, because they were so—you know, we were working in 8 bits and memory that was measured in megabytes back in those days, right? [Laughter] So, I learned how to be an assembly language programmer to push the systems to their edge and build fun games and all that kinda stuff.

Brian Dawson: I love it. Well, my history in the console industry—so, I first, when I decided I wanted to code, started with BASIC. I was forced through Pascal, but did some cool things with it. And then immediately dropped down into assembly language—

Robin Purohit: Yeah.

Brian Dawson: - and fell in love with it. I still think that procedural C is awesome, but some people may think I’m crazy.

Robin Purohit: Well, I think, because you do assembly language, can. you do math and hexadecimal?

Brian Dawson: You know, you’re gonna make me feel old now. Because of the story I told you about my career trajectory, no longer, but I used to think in hexadecimal for a period of my life, there.

Robin Purohit: It’s a good party trick. [Laughter]

Brian Dawson: It sounds like—can you, I guess, is the question?

Robin Purohit: Yeah, I used to be able to do multiplication in hexadecimal.

Brian Dawson: Oh, okay. I wanted to challenge you. That would be a good thing for the episode. So, I guess jumping ahead into—so, thank you for sharing that, but jumping into more of the now and the future, like where you’re living today, you recently shared some 2021 predictions. Peritus actually had a pretty significant footprint at the Predict 2021 conference if I’m recalling correctly, and I think associated with that, you have also shared predictions, and you said that you believe AI will play a much bigger role this year, not in the future, in helping technical people. And I’m assuming it speaks to reducing cognitive load, improving quality, but I’d like to hear—what do you mean? How do you believe it will help technical people and how this year, what are we gonna see?

Robin Purohit: Yeah, I think—well, one of the side effects for the tech industry from this COVID thing that we’re all living through and will live through for many years, I think, is a huge appetite to move faster to digital services and the cloud.

And the level of technology sophistication you’ve had to have to take advantage of open source and cloud technologies and now data science is—there’s so much change happening so quickly, and there’s so much demand to build these cool things and transform businesses to become digitally native that there’s a gap. There’s a gap between the people that are, you know, Silicon Valley rock stars that can do it and the rest of the community that wants to be like them, right?

Brian Dawson: Right.

Robin Purohit: So, that’s kind of the big problem statement that I think is creating the demand. And the good news is that there is a lot of automation and collaboration that is now happening among the development community. So, if you go to a placed like GitHub or Stack Overflow, you'll see just a huge volume of hundreds of millions of engineers collaborating out in the open.

Robin Purohit: So, that provides a rich opportunity to provide AI and machine learning to help them find help on solving a problem, building a new product, learning a new technology.

Brian Dawson: Awesome. And how much—I guess I was gonna ________, there was a, look, if we look back when you were 16 and software was this black box that you just found, and I like to joke you found the smart guy that did computers and you asked them to build a web app for to build an application, I’d like to say that working in a silo or working in a small team, you weren’t automatically burdened with the extra layers of coordination required to work with larger teams, the extra layers of checks, balances, and coordination that are required when your organization realizes they care about what you’re delivering and they need to know how it’s delivered and they're investing IP in it and it needs to be maintained.

So, I posit that as software is recognized as being more important to business, so does that cognitive load that is being placed on individual developers.

Robin Purohit: Yeah, yeah.

Brian Dawson: And that part of what we can do with AI is automate away that rote stuff and free up some cognitive capacity to do better things and, to an extent, I’d say make software development fun again.

Robin Purohit: That’s right.

Brian Dawson: I’m curious if you have any comments on that or whether that relates to your view of helping technical people.

Robin Purohit: That’s right. I think, you know, especially, I’ve lived all of my business career in the last 20 years in business critical software, and now, almost all software is business critical, right? [Laughter]

Brian Dawson: Yes.

Robin Purohit: I think unless you’re a consumer company, there’s move fast and you kinda have to move fast and make things work, you know, that’s kinda [Cross talk]. [Laughter]

Robin Purohit: And there’s a lot of things to be aware of. It’s not just running good code, but it’s, you know, how does it integrate, is it gonna be secure enough, is it gonna be resilient, how does it integrate with everything else? There’s so many things to be aware of that honestly, I think it just outstrips the average developer. There’s a lot of amazing ones out there, but for the average developer, it’s very tough.

Robin Purohit: Even harder for people that are on the other side. So, now you roll it out, and if you’re a poor software support person who never wrote the code, now you’re responsible for supporting code that might have 100 different components, 1,000 different components authored by many different people. Some are open source, so there’s nobody to talk to, and you’re on the hook if something’s not working. So, how do you solve that problem, right? That’s even, I think, bigger than the developer problem, quite frankly, and more business critical.

Brian Dawson: Well, yeah, it’s interesting, and I’ll get to the specific questions, I’ll squeeze in some commentary, here, that maybe we can talk around. But I even think about—look, when source control, for those people that didn’t start in mainframe, when you didn’t really have central source control servers, when all of your version control was done mostly locally and then you handed a set of code off to somebody else in another cubicle or down the hall to figure out how to make your stuff fit with somebody else’s. And then what happened, you’ve got some big binary that got dropped on a server somewhere to deploy waterfall to the umpteenth degree, it was easier to kinda sit in your silo and not think about the macro picture, but, you know, with the realization that there’s flaws in developing that way.

With the emergence of continuous integration, continuous delivery, and DevOps practices, it feels like, while we’ve begun to streamline the process, the fact that the scope of the stakeholders—look, we’re moving faster now. Marketing has to be able to market it fast, Sales has to be able to sell it fast, and if it’s not SaaS, Services has to be able to install it, Support has to be able to support it. So, even as we gain or optimize in one area—and I think it’s all moving in a positive direction—the rise of importance of software, the expansion of other stakeholders or bringing other stakeholders into it, to your point, and I’m probably just using a lot of words to repeat what you said, right, places a little bit more burden on the developer.

Robin Purohit: Yeah.

Brian Dawson: So, yeah, I think it’s awesome that you are focusing on that. This is gonna lead me to say, before we get into the downstream support, when you look at Peritus focusing on AI and you look at your experience over the arc, I think, of modern software development, what’s the convergence, what convergence, if any, is there of the CI/CD and DevOps practices or DevOps approaches and artificial intelligence?

Robin Purohit: Well, you know, artificial intelligence is about making people productive and being able to allow them to make the decisions can be very complicated in a much easier way, right? So, an average person can make an expert decision.

Brian Dawson: Oh, quite right.

Robin Purohit: If you think of the CI/CD process, there’s a lot of high judgment decisions that have to be made on whether the software is good enough to be integrated, whether the test results are good enough to roll out. And then, you know, if you fill this chain if you’re a site reliability engineer or whether the problem you’re seeing can be fixed and what is the best way to fix it.

So, those are very high cognitive load tasks, to use your word. And so, the great news is that with all the automation and collaboration that underpins CI/CD and DevOps, there’s a lot of sources of insights and a lot of repeatability that you can use with machine learning to allow those people to make better decisions. So, that’s really the place we’re soaking in.

I would say that while we’ve been talking a lot about developers—in fact, the tech support arena is probably where there’s the biggest pain point right now. Because they're the—you know, when I worked at Mercury, we used to have a saying that said, the developers have the party and the Ops guys have the headache, the hangover. [Laughter] And when you have a hangover, you’re very motivated to find something that can help.

So, it’s one of the reasons that, early on in our company, we’re focused in bringing our AI recognition engine to the world of tech support. Because if you’re in tech support and you have to open a ticket with a developer, you’re gonna need help based on what they’ve done, you know, how they took something in the past. It’s gotta be a completely different paradigm than the old take a ticket and wait in line process, right?

Brian Dawson: Right.

Robin Purohit: You know, that’s why we’re focusing on that problem first.

Brian Dawson: So, let’s dig in a bit more specifically, and as a matter of fact, actually, I’ll shift, but maybe you can cover more specifically how AI can enable better support and what it means for the overall software discipline.

But before we go there as a branch off point, I’d like to call out an article that I was given to read that you posted on DevOps.com on the use of community forum for technical support, starting out and referencing Stack Overflow. Based on that, is the difference between best in class versus best effort forms and can you help explain AI’s role in that problem we’re trying to service with those forums?

Robin Purohit: Yeah, the heart of that article that I did was about looking at community forums as potentially the best place to answer technical questions, right, all out in the open. Because we fundamentally believe the old world support is all gonna move to this kinda more open collaboration to solve problems rather than kind of, you know, take a ticket behind a firewall.

Stack Overflow has been kind of the community that’s created the wave of the future. There’s over 125,000,000 active users a month, most of which are developers or full stack engineers, looking for answers to problems related to open source and cloud technology, right? A vibrant community with a lot of topics.

So, what we want to do is understand what’s the state of play? You know, because we help look at around 50 different community forums, some which are hosted by open source companies, we looked at Stack Overflow forums as well as ones hosted by an IT vendor, a traditional IT vendor. We try to understand—well, how effective are those forums for helping technical people to get answers to their questions, right?

And what we found was, on average—so, what we call best effort—is that about a third of the time, you can get an answer to your question and much less than, I think, 20 percent of the time, you can get an answer to your question in the first day that is acceptable. The best ones, and these are generally only—

Brian Dawson: Just to make sure I captured it—20 percent of the time, acceptable answer on the first day, did I hear that right?

Robin Purohit: That’s right, exactly.

Brian Dawson: Okay, sorry to interrupt.

Robin Purohit: And it’s important just to think about that, because if you don’t get an answer to your question your first day, and it’s something important, and usually it is, you’re more likely to open a ticket or escalate somehow, right? And so, that, to us, is a really important SLA, if you want to call it that.

The ones that do the best have substantially better performance, generally two times better. So, take your open core communities or open source communities with a lot of people trying to help each other or build a reputation by answering questions, you'll see about double that. So, 40 percent of the questions on the best ones get answered their first day. We think it could be better, but that’s a pretty good benchmark, and the majority of the questions are getting answered.

Probably more strikingly is the participation in answering questions. So, on average, for what we call best effort, 20 percent of the people in a community are actually generating the answers—only 20 percent, right? Whereas on the best, it’s almost flipped around, so it’s generally 60, 70 percent or more are all participating and answering the questions that are there. So, it serves as a signal of an active, engaged community around a particular topic.

Brian Dawson: So, I’m interested—and I may be jumping on your storyline, but wouldn’t some say that that’s just, some people are better at community management, or is there more to it?

Robin Purohit: Yeah, I mean, certainly, that’s part of it. You have to create a community with an identity and with the right expertise that’s engaged on that community to make it an effective place to get answers, so I think that’s kind of pay to play.

Robin Purohit: But I think it’s really about having the intrinsic motivations that makes—it’s valuable for me as a person to go to a community to answer a question. I’m gonna get some love from the community back, but even better if I am more hirable for my next great gig because I’m known as an expert in a really hot area, right, or an area that’s really important to the IT industry.

So, to go back into AI, so, we took a look at that to understand what’s the nature of the problem, who does well, who doesn’t do well, and so, where can we take the industry? So, we’re looking at a variety of things that an ML can do to understand all of the prior conversations that happened out in public, all of the related content that’s kind of trusted content on the Internet, and to help people answer questions much faster. We think that’s a huge opportunity. And by doing that, you boost somebody’s ability to look good on a forum and improve their career prospects and their industry rep and be more employable.

Brian Dawson: So, I’m gonna ask just a bit of a devil’s advocate question, just to start a little trouble here, Robin.

Robin Purohit: [Laughter]

Brian Dawson: That’s awesome, right? And bear with me while I play the jerk, even though I’m probably too good at it, right—but okay, that’s awesome. A person on Stack Overflow or a related community forum can answer a question better, and we also know over the past decade, I think something to call out is, and maybe most of our listeners know—there was a time where very few people had community forums, and really, over the past 10 to 20 years, it has become almost standard for every tech company to have some level of community support.

And great, we got it out there, but what does that matter to me as someone early in the software development life cycle? What does that matter to me as a business? Why am I concerned about somebody answering a question better and AI helping them do that?

Robin Purohit: Yeah, two things. One, if you’re a developer and you’re trying to stay on the cutting edge of technology, you wanna know you can go to a place where you can get an answer quickly, right? So, I think you wanna be the beneficiary of, you know, AI augmented people that are empowered to answer your question much faster than before and without having to open a ticket.

And then if you flip it around to the other side, you know, if you are a developer that released an awesome product, and usually you’re onto your next project right after that, the last thing you wanna hear is a call from Support saying, “We have a problem, can you help me diagnose this?” Right? So, the better you can get a handing off what you’ve done with all its learnings and iterations so that a support team can be more self-sufficient, the more time you can spend on innovation.

Brian Dawson: I love it, and if I may propose a couple of other things, right, that I think fit is, you know, one is, we talk a lot about this practice of software delivery management. We’ve invoked the old adage that if a tree falls in the forest and no one’s there to hear it, does it make a sound, right? And at the end of the day, I believe there’s an emerging belief in the community and the industry and I would assume you and Peritus are aligned with it, that delivering software is just developing code and putting it on a server, right? Delivering software, that’s the tree falling in the forest, right? At some point, you have to make sure somebody hears it, and hear it isn’t just, “I downloaded it,” right, and it’s not just, “I opened it up,” but, “I opened it up and I used it to accomplish what I was trying to accomplish and, to get corny, I find delight and joy from it.” So, I would contend that oftentimes, your community forums are one of your primary interfaces with the people that you are trying to solve problems for.

Robin Purohit: Yep. There’s a social role now, right? So, especially anybody who’s under 40, they're used to discussing things in the open. That’s kind of their mentality and they don’t understand why you don’t wanna do that. So, I think all the consumers of your product are gonna force you to have a kickass community forum that answers questions. [Laughter]

Now, I will say, there’s a lot of different kinds of community forums, as we’re talking about it. The thing that I think is really important is, a lot of vendors have tried to turn their community forums into marketing machines. There’s nothing wrong with that, but I think what sometimes they do is, they conflate a community with its intent. There needs to be a place to go where you’re gonna get technical help and expertise from the best people on that topic and your product, and then there’s other things where you’re trying to connect with and socialize and understand and know what the next big thing is. I think it’s super important that vendors don’t conflate those two things.

Stack Overflow, as we talked about before, does a really good job and it’s so popular because it is solely focused on helping you find an answer to a technical question, and the moderators make sure of that. So, that’s why I think it’s been so successful for that purpose.

Brian Dawson: Yeah, a couple of other tangents, we’ll make a left and a right and then we’ll come back and then we’ll land on what we agreed to talk about.

Robin Purohit: [Laughter]

Brian Dawson: But we did some studies, I have done some studies over the past couple of years talking to over 50 developers, asking them what are their motivations, frustrations, and fears. And a range of brand new developers that call themselves brand new full stack developers, medium level, high level, and think about what are a lot of these purveyors of these support forums if you have an offering that is developer supported or developer targeted. And one of the interesting, biggest fears was being perceived as an impostor, right?

Robin Purohit: Yeah, that’s right. [Laughter]

Brian Dawson: And that paired along with being perceived as an impostor, but also having to disrupt my colleagues and interrupt my colleagues to ask them to clarify something for me.

Robin Purohit: That’s right.

Brian Dawson: And so, what they really wanted to be able to do was have a quick way to shore up their expertise without having to interrupt anybody. But there’s another thing that ties to what you’re talking about here, Robin—the other kind of motivation that was there for developers is, I like being perceived as an expert. I like helping people. So, if I’ve collected the knowledge, I wanna kinda be able to use it to help people and be celebrated for that. Any thoughts on that?

Robin Purohit: No, I think that’s right. So, two things I did pick up on. One is, our strong belief is that building AI systems that help people is the better way, is really what AI should be trying to do rather than solely trying to automate their way out of a job, right? Because I think technology is moving fast, it’s getting more complex, augmenting their ability to do more and more is really, I think, a better mission for the AI industry.

Robin Purohit: But you’re right, if you wanna go to kind of a safe place to get some expert insight and then they wanted to test their ability to use that insight out in public, so that they look better.

Robin Purohit: That’s the fundamental human motivation, but also, you know, the tech world is a highly meritocracy world, so if you say something in public and it’s not right, you'll probably get pounced on. If you say something that’s insightful, especially if you pounce on something quickly, you’re gonna be a hero. So, we’re trying to get more people to look like heroes and be unafraid of participating in these public—

Brian Dawson: I love it. And I think, and to that end, when people are participating more, when they feel like heroes, their connection to your product offering—

Robin Purohit: That’s right.

Brian Dawson: - your community is stronger. And when you have a stronger customer base, you have happier and stickier customers, you have a better business.

Robin Purohit: More reference-ability, yeah, more ideas coming your way. That’s exactly right.

Brian Dawson: Yeah, that is—you know, honestly, a way that I had not thought about, this goes into the bucket of really novel yet impactful ways of applying AI that I hadn’t thought about, hadn’t really dawned on me, but are still core to this whole process of delivery in software. I can’t wait to see more.

I did want to share an analogy, and then I want to ask you for some tips for companies in managing their forum. But I really—so, I had to put in hardwood floors this weekend and I’m older than I wanna say, my back hurts, my hands hurt, and I was just telling my wife, “I can’t wait ‘til those military exoskeletons that they're working on to make their super soldier, ‘til I can have those to build floors,” right?

Robin Purohit: [Laughter] Yeah.

Brian Dawson: But if you think about it, the way you phrase AI, it’s not about building robots for people, right? It’s about building exoskeletons to multiply whatever their abilities are, right?

Robin Purohit: Totally. Yeah, that’s right, that’s right. I think that’s exactly the way we ought to think about it. It’s also, to me as a humanist, it’s more motivational, right? I mean, there’s always going to be automation, that’s an inevitable outcome of great technology, but I think it’s far more motivating getting up every day with your team saying, “We’re gonna help people be smarter, more effective, more productive, be happier in their work.” You know, I can get up every day and work 18 hours a day working on that problem versus getting up and saying, “I need to take a million jobs,” right? That’s far more motivating to me. [Laughter]

Brian Dawson: Excellent. Right, yeah—great point, love it, love it. So, let me ask, going back, pulling back in—so, how can companies improve? What are tips for companies, including and/or outside of AI? You mentioned, don’t try to make your community a marketing vehicle, or at least not covertly, directly, and grossly. Any other tips that you have for companies interested in building communities?

Robin Purohit: Yeah. The main thing and my last point is just to segregate it out, make it very clear what parts of your community forum are for a marketing space and for networking and socialization and all that stuff, but make sure there’s a specific place to go to get your technical questions addressed, right? So, I think that’s a best practice.

The second thing is, you really wanna get the best people participating. Like, the secret to making a thriving community that takes on more questions and ultimately leads to less cases coming in and distracting your team is having the best people participating. And those best people could be from your customer base, from your services ecosystem, your engineering community. But you want to encourage everybody to actively jump on and mobilize to answer people’s questions and help each other out. And that’s why we’re—you know, what we’re trying to do is make that easier with some AI capabilities to help everybody do that better, so that everybody has the best information possible at their fingertips for the nature of what is being discussed. That’s, to me, a way to bring AI to it.

The other thing to focus on is, where is the business driver for community forums if you’re a vendor, right? So, there’s probably two kinds of communities to talk about. One is if you’re a vendor, and a big goal for a community forum if you’re a vendor is deflecting cases. You’d rather have a question and answer ultimately on Google search and if not Google search, people can go to your forum and they get the right answer there. And it should be a last resort when somebody opens a ticket. And, you know, the cost savings are enormous. The average tech vendors, the cost for ever open case that’s not a trivial case is usually a few hundred dollars.

Brian Dawson: Wow.

Robin Purohit: And we will get—you know, the biggest vendors get millions of those cases a year. Even an average sized vendor, a mid-sized tech company will get hundreds of thousands of those. You can do the math. You know, that’s not where you want to be spending your money. You want your technical service experts to be working on the hardest problems, right, not the everyday problems.

Brian Dawson: Yeah. That sounds like, assuming that estimate, which, when you scale it up is really impactful, but I assume that’s not accounting for the cases where it involves remediation or bug fixes and then testing that. That can [Cross talk]—yeah.

Robin Purohit: Yeah. Sometimes, you need a service person to do something for you, but you know, you want it to actually be when they're taking action on their remediation stuff.

And even there, there’s more and more automation that can be put in place to do that remotely with telemetry and now with software, a lot of things can be rolled out in a distributed fashion. So, how do you take the right action is another one of those outcomes of using machine learning technologies to tackle these problems, yeah.

Brian Dawson: Awesome. Well, thank you. So, I’m gonna shift into some of our standard questions here. I don't know if you’ve heard them before, but at the end of every episode, we ask about DevOoops. That’s not mispronounced DevOps, that is Dev-O-O-O-P-S. I guess it’s more DevOoops!

Robin Purohit: Yeah.

Brian Dawson: And this is an example of a challenge, it’s a technical challenge usually, or in the technical space that you face during your career that, ideally, just for entertainment is almost embarrassing to talk about, but you learned a lot from it and you can help our listeners think about how to avoid that mistake. Do you have a DevOoops you’re comfortable sharing with us?

Robin Purohit: Yeah, well, I think the DevOoops is a common thing across many companies and teams I’ve lead over the last 10 years. And that said, not all clouds are built the same, right? And it used to be, we talked a lot about virtual infrastructure , now we’re talking about container infrastructure. It is really hard, you know, to figure out how to deploy your software on the latest and greatest containerized or virtualized infrastructure and have predictable outcomes.

There’s so many differences in the various, even, container orchestration where it works out their nuances on how the underlying containers or virtualized infrastructure operates. And most companies, every product I’ve had that’s a cloud product, runs on multiple clouds, some on premise clouds that are virtual clouds, some converted, some by a big public cloud vendor, by a hosted cloud. You kinda—usually, as you grow, you’re forced to deploy in many different environments.

Robin Purohit: Yeah, no, it’s hard. So, I’ve had, like, my best crackerjack cloud infra developers in the world working for me and I see them struggle because the technology is moving so fast and there’s so many variances. So, to us, that’s an opportunity, right? How do you help, just looking at this cloud native transition which is obsessing most of the Dev community now, how do you help them understand those variances and still be successful?

So that they're as prepared as possible for all the variances and the different environments and can tune and tweak what they're doing.

Brian Dawson: Yeah, can I ask—and I know you’re in a precarious position, right? You’re the CEO and Co-founder of a key business, here—but do you have a case that you can share where failure to recognize either yourself or a developer in planning the difference in clouds has resulted in a pain?

Robin Purohit: Yeah, I think, going back to my time at BMC, we put a lot of our software on all these different kinds of clouds. And so, I remember once, we did a—we had a rollout to Amazon and after running for many years primarily these kind of hosted cloud environments run by SIs, and there’s two things that were a really big ooops.

One is, the performance was completely different than what we expected, right, because the dynamic virtual infrastructure just behaved very differently than the dedicated ones, dedicated private cloud stuff we had, and second, the build was just unbelievably high.

Robin Purohit: And especially if you have any application—and it’s getting better now, but if you have any application that uses persistent storage and persistent databases, the cost of the cloud is actually more expensive than doing it on premise. So, we had to learn a lot of lessons very quickly in when we—what and when do we wanna put stuff on a public cloud, and then how do we re-cue our service to be more cloud native so it didn’t incur these sticker shock prices. So, don’t put your old stuff on the new stuff is probably the best advice I could give you if you want to save costs. [Laughter]

Brian Dawson: Yes, no, thanks for sharing that. I think for the people that have been burned that have learned lift and shift just doesn’t work. While I don’t want to confuse correlation with causation, it is—you know, there’s various reasons, people, so don’t pick on me, but it is, look, the word ephemeral versus persistent, I bet if I was to go run a scan on where the popularity peaked, it came right along when people realized their cloud spend was blowing up, because they just picked up apps they built on on prem environments that they paid an electric bill for, right?

Robin Purohit: Or a lot of highly motivated consultants showed up and said, “We can take you to the cloud” and most CIOs were being asked years ago, “Why aren’t you on the cloud?” So, you know, [Laughter] it was kind of a while before stuff was ready. So, that’s why I think they're seeing this huge interest in cloud native architectures and how do you go faster to that world. Because you, you know, people don’t realize, you’ve gotta rethink your software so that it’s gonna be really optimized to take advantage of these new, both high performance and heightened ultra-scale capabilities that can be cost effective if you do it right, yeah.

Brian Dawson: Well, since I really squeezed you on the DevOoops, I’m gonna squeeze you on another one—Canada or the Bay Area? No, you don’t have to answer that. Which is better? But no—

Robin Purohit: Well, I’d say, the sad thing is that, so, one thing we did, I also had an aspiration to open a big Dev center in Canada. So, post-COVID, we decided to accelerate all of that. So, I’m now building up, the core of my designs team is all in Canada now. It’s a great hotbed of AI and ML talent, and so, we decided to put our stakes down deep up there.

Brian Dawson: Ooh, stake your claim early before everybody—

Robin Purohit: No, it is crowded, but it’s a thriving ecosystem, so.

Brian Dawson: Where at? What area of Canada?

Robin Purohit: Our virtual headquarters is Montreal.

Brian Dawson: Okay, nice, nice. Thanks for sharing that tip. Hopefully, we haven’t tipped off competitors to rush into Montreal.

And just so everybody knows why did I ask, because Robin is originally from Canada, he now resides in the Bay Area, and I’m a Bay Area native who happens to love Canada, but I don’t want to put him on the spot to pick a favorite.

So, I’ll shift. We also ask about key resources. What is a book, podcast, or other resource—blog, person to follow on Twitter—that you absolutely recommend our audience consume? And it doesn’t have to be technical, but something that has been informative and impactful for you.

Robin Purohit: So, I’m a big fan of reading non-technical things and non-business things. I think you can learn so much from literature or reading about other fields. They give you the metaphors to figure out what that next thing is, right? And I think that’s been a source of inspiration for a lot of tech entrepreneurs.

So, one that I would recommend that people read, especially since we’ve been talking about AI, is The Secret Life of Trees. It’s a great book that was published—and there’s been a movie, there’s a movie on Netflix if you’re not a big reader called Intelligent Trees that explains it quite well. But basically, it goes into this hidden neural network that sits underground in a forest—because we talked about the forest and the trees before [Laughter]—that allows trees to communicate and take care of each other and collaborate through this super diverse, basically, neural network. And so, it’s very humbling to see something organic that works that way that we never would’ve expected. I think there’s a lot we can learn from it.

Brian Dawson: Oh, thank you. That is gonna go up—you know, every once in a while, and no slight at previous guests, right, there’s the common core, the staple books in our technical space, right? Phoenix Project, Accelerate—all phenomenal, and if you believe…but then I always, it strikes me when people get an unexpected book, like The Alchemist, another one of our guests suggested. Secret Life of Trees, that really—I’m excited. Early in my years at PlayStation and in gaming, I was very invested and interested in neural networks and implementing neural networks, so this is especially intriguing to me. So, thank you for sharing that. That is one that I’ll capture and share out and hopefully read soon.

Before we get ready to wrap, Robin, any final thoughts for our listeners?

Robin Purohit: No, I think just to keep on top of the new tools and techniques that you’re gonna have that are fueled by machine learning this year. If you’re a techie, there’s going to be an army of new offerings that will help you do your job better, whether you’re a developer or a support person, in operations. We’re not alone in sort of thinking about this problem. I think that’s a good thing, because there’s more than enough opportunity for a lot of successful startups, but keep your eyes on all these publications and podcasts like what you’re doing because there’s a lot coming.

Brian Dawson: Awesome. Well, thank you, I will do, and I personally will say, I am excited for, I think, machine learning and artificial intelligence becoming real. I do have a little bit of fear and anxiety, but the conversation with you has made me more excited. Robin, thank you. It really was a pleasure to get a chance to talk with you. Again, I highlight or the team, as you listen to this podcast and look into Robin and his new venture with Peritus, just soak in his arc of experience in the industry and Robin, it was a great opportunity to get your perspective on these things. Thank.

Robin Purohit: Thanks, Brian. Thanks, again, for having me today.

Brian Dawson: Alright. Thank you for listening to this episode of DevOps Radio. I hope you enjoyed it. Please tune in for our next episode.