Adam Arakelian at Dell talks machine learning and AIOps to accelerate software development pipelines

Adam Arakelian, Senior Director of Engineering at Dell, returns to the podcast with the inside scoop on how his teams are utilizing machine learning and AIOps from a DevOps perspective. Discover how to improve your developers’ experience by removing toil and friction and find out where ML and AI intersect at Dell with intelligent chatbots.

We discuss Adam’s Tricentis Virtual Summit session. Watch the session on-demand.

Listen now 

Spotify | Apple Podcasts | Buzzsprout


Watch episode

Podcast transcription

This transcript has been edited lightly for clarity and brevity.

Emma: Hello listeners, I’m very happy to introduce our guest, Adam Arakelian, Senior Director of Engineering at Dell—our second guest to be invited back on the podcast and held in high esteem. To remind our listeners, you’ve been directing engineering at Dell EMC for around seven years.

Dell hardly needs an introduction, but as you said when we last spoke, Dell has a myriad range of products beyond just laptops. You’re delivering software as service on-premise and embedded hardware. Your fortes lie in software leadership engineering across all aspects of product development, and that’s from building products to driving DevOps. Operationally you’re managing over 50 teams, so it’s a large scope.

Let’s start with the initiatives that you shed light on the last time around tools centralization, continuous integration with your philosophical continuous acts, DevOps framework, and engineering agility. How are these projects going, and how are they helping your team deliver quality?

Adam: They are going really well. What this ultimately turned out to be was a cultural and philosophical shift. It really manifested itself in a platform that helps us facilitate how we work around driving quality and compliance and security left, measuring CI/CD maturity and Agile maturity across our Dell ISG organization. By shifting quality compliance and security left, we’re able to ensure a level of consistency. It’s that consistency that’s so critically important to software delivery.

And it also allows us to ensure a level of maturity that our product groups need to meet. They’re heavily based on DORA, the Google DevOps metrics, but we’ve slightly modified them because we’re not all cloud-based products; we have on-prem and embedded hardware.

“We’ve made those metrics fit our needs so that we can drive towards the same low, medium, high, and elite measure of maturity. The focus is to put our products in a position where they’re at that elite status all of the time, so that we can ensure the highest level of quality, compliance, and security for our customers.”

Emma: Marvelous; that quality and high-tiered premium service feeds into continuous X—I really like that it’s an underlying philosophy, and you’re using DORA to give you really good visibility into metrics. Those metrics are real-time and continuously being fed into the method, continuously adapting it.

Adam: Yeah! It’s good to use those as industry standards, so we understand where we’re measuring up against the industry overall, and we’re very much driving security in the way that we produce our products. We’ve all heard about the recent hacks that have happened over the past year or so with regard to open source.

Our focus has really been on vulnerability management. We’ve put a concerted effort in shifting that very, very left, and we’ve certainly driven that across this platform.

Emma: Awesome! Catch the bugs early and then you’re a lot less vulnerable to cyber-attacks and whatnot.

Earlier this year, Michael Dell said that all of those products and services are available on a consumption basis, and in a very developer-friendly way, and that’s a big priority for Dell this year.

With this focus on DevOps strategies from the top, how are you executing? I know that you’re a big fan of philosophy over method, but I’m interested in how this is playing out for you and your team.

Adam: Really well! It’s a strategy that Michael Dell has been talking about and it’s been publicly announced. The focus is on our developers who are using our products, whether they’re cloud on-prem or embedded, and we’re driving towards what we’re calling sort of that developer-centric experience.

Late last year, we published developer.dell.com, which is our centerpiece for ways that our customer engineers can get information about how to use our products and leverage APIs. We’ll be adding more and more content to that over this year and going into next year. It comes from an understanding that the developers that are using our products are focused on building products and workflows, rather than focusing on infrastructure.

“We’ve come to a point where we realize infrastructure is a commodity, it’s a utility—it’s no different from electricity, oil, and gas—it has to be there; we expect it to be there. This is the advent of the cloud, right? Developers need infrastructure; it’s a necessity, but we want to make it as easy as possible for Dell products to fit into our customer’s workflows, and help their developers facilitate the building of their products.“

Adam: Our developer.dell.com site gives us that centerpiece for developers to be able to get all the information they need in order to be able to do that, so we’re driving this strategy. We’re early on in the initial phases of it, but over the next several years we’ll be focusing on it more and more.

Emma: Great. I checked out developer.dell.com and it’s a huge catalog of resources, with APIs ready to be lifted and suited to both cloud and on-prem.

The reason we got back in touch is because you’re one of the first people who agreed to run a session at the upcoming Tricentis Virtual Summit. This felt like Christmas for us because the event is all about putting the focus on what our customers and partners are up to, and how you deliver innovation with confidence.

You’ll be running a Tricentis Virtual Summit session on machine learning and how you’re using AIOps to accelerate your software development pipeline. In a nutshell, what will be you be covering in this session?

Adam: Yeah!

“There’s some really awesome stuff that we’re beginning to do with machine learning and AIOps from a DevOps perspective. Now that we’ve built out our platform for developers to begin to leverage for building product, we’re collecting a ton of data and information. And all of that data and information can be used to do some incredible things to remove toil and friction that hits our developers in their day-to-day lives.“

Adam: There are heavy intersections between things like ML and AI, for example with intelligent chatbots. Part of removing toil and friction in developers’ day-to-day lives is in helping them understand what might be wrong in their pipeline, instead of having to troubleshoot it themselves or contact an SME to better understand. For example, in product engineering, to have developers look at a problem. Imagine being notified through a platform that there is a problem—with suggestions on how to fix it—potentially without having to do very much.

What gives us an ability to reduce our time to restore? Time to restore isn’t just about getting infrastructure back up and running. What I like to look at are things like: how do I get back to my normal working environment? What do I need to do to get that back up and running, working and functioning so that I can be at my most productive? Having that intelligence where we can better understand algorithmically—because of all this data that we’ve collected—what might be the problem and how a developer or productivity engineer can fix that problem, helps tremendously. That’s one major way that we’re beginning to use this data from an ML/AI perspective.

Emma: When you originally shared with us your session description, the statement around determining what is abnormal about a CI/CD pipeline using intelligent chatbots, was clearly born out of experience by doing so in your team.

I guess incorporating ML/AI was the natural progression seeing as you had all this data, so have you seen really solid results to reduce toil and whatnot?

Adam: Absolutely. It starts with as simple a thing as looking at things like build time.

The way that typically people become informed their build time has increased is after days or even weeks. In some cases it’s a developer saying, ‘Hey, my build time is increasing, what’s going on?’ Those might be slight, maybe by two or three minutes; it doesn’t have to be six hours. But when you’re used to a five-minute build time or a 10-minute build time, and you get a 20% increase, that’s where we begin to see a set of behaviors that are outside of your scope of abnormality.

That goes both ways, right? If we have a decline, then why did it decline? If I’m used to this perceived exertion of build time and this level of effort that we’re putting in on a consistent basis, and I begin to go out of that range, the question is: why?

“I don’t want to have to spend hours or days figuring out what’s going on. I want to know in minutes—ideally, in seconds—what happened, and what we need to do to address that. And so, simple applicability of ML/AI to inform build time can help other developers increase their experience and productivity; and get them back on track.“

Emma: Those seconds add up to hours in the end; If you can have that real-time feedback then you’re only going to get better at delivering quality software faster.

In 10 words or less, could you give some advice for anyone looking to incorporate machine learning AIOps into their software delivery?

Adam: One of the things that we struggled with—which I’m sure other people struggle with—is data management.

“Data management is key. Really understanding where and how your data is managed is super critical. Where are you getting your data from? How are you managing it? How are you bringing it together and coalescing it? That’s absolutely critical to driving any kind of ML/AI experience.“

Let’s finish with a look to the future. Why should our listeners be sure to check out your session at the Tricentis Virtual Summit?

Adam: We’re going to talk about ML and AI and the use cases around them.

“We’ll dive into some of what we’re doing at Dell and really talk about how we are improving our developer experience by using ML and AI. We’ll talk about things like chatbots, predictive analysis, alerting, thresholding, and how we’re embracing this in our DevOps teams.“

Emma: There’s nothing better than seeing real-life examples, especially for a company like Dell with you spearheading that innovation. The Tricentis Virtual Summit’s theme is ‘Delivering innovation with confidence,’ which really is the name of the game for you and your team at Dell.

It’s really awesome to see Tricentis customers and partners like you on the cutting edge of ingenuity. You are a real partner in driving and shaping our industry together.

Check out the latest podcast episodes for more insights from thought leaders like Adam.