Listen
Links
Subscribe
Transcript
[00:11] Tim: Welcome to episode 3 of the Into the Hopper podcast. On this episode, I’d like to welcome Mr. Mike Rogers, the Senior Director of Solutions Architecture at Tala Security. Way back in the year 2017, before coronavirus. Mike was my manager on the Solutions Engineering and Implementation team at Distil Networks for several months, and we had a great time working on some data science problems as they applied to solutions engineering during that time. And Mike is a big believer in data science and analytics and offered to come talk to me about it from his perspective outside of data science and analytics and talk from his experience with customers as to why this is such an important and valuable topic. Welcome, Mr. Mike Rogers.
[00:59] Mike: Tim, thanks for having me.
[01:01] Tim: I’m delighted. You’re one of my favorite managers I’ve ever had, and I think that’s probably because it was only for like 2 months. It didn’t last long enough for—
[01:09] Mike: Yeah, makes sense. I’m easy to get along with in short bursts.
[01:13] Tim: So what is solutions engineering and/or solutions architecture? I don’t even know if they’re different. What do you do?
[01:20] Mike: Day to day, what we’re trying to do is we’re listening to customers, we’re listening to the product team, we’re listening to the vision of our business, and we’re trying to match all of that together and come up with solutions to problems that customers have. In my role at Tala, I’ll give a quick kind of thumbnail on what Tala Security is. Tala is a solution for protecting the data that you’re putting in the browser, really building an invisible fence around all of the software, the JavaScript and whatever code is running in the browser from that you’ve developed, that third parties have developed. If you’re relying upon frameworks and open source, whatever that is, this kind of idea of a web application runtime protection system that governs and ensures that things don’t go rogue and start stealing data and taking it off. Commonly, you’ll hear this described as kind of the Magecart tactics, but we’re seeing a lot more and more beyond just those quote-unquote Magecart groups. This is happening now across all industries, not just e-commerce.
[02:22] Tim: The solutions engineering team is helping get that product in place with your customers.
[02:29] Mike: Yeah, I mean, we get into a conversation, we’re doing a little bit of evangelizing and talking about this problem and educating, and then You know, when customers who are aware that they are facing this as a challenge, maybe it’s a gap in their data protection strategy, you know, we go and look at their infrastructure. We think about the ways that our software and solution can integrate there and how we’re going to demonstrate value to them, make sure that they have all the, you know, everything that they need to solve these kinds of problems, protect the data, protect their customers, you know, avoid the compliance concerns, that sort of stuff. So we, you know, sales, a little bit of architecture and engineering, and a lot of kind of customer success and customer journey, you know, table setting.
[03:11] Tim: Your background is mostly in security-type products, is that correct?
[03:16] Mike: Yeah, I would say in the last decade, that’s absolutely true. I worked at Neustar and at Distil Networks for 6 years. I was part of the early team there, the 13th employee, I’d say. We grew that business, and most recently I’m here with Tala.
[03:37] Tim: In my experience, I’ve been in several different security companies. Modern cybersecurity products are generating enormous amounts of data.
[03:47] Mike: Absolutely. And it’s, you know, it’s an area that I think also is ripe for opportunity, right? If you look at the marketplace around things like security incident and event management platforms, the Splunks and homegrown-type solutions that you might see there, or even the data lake technologies that exist. You know, everyone’s looking at ways to dump that data in and save it for whatever that they’re saving it for, at least on the customer side. And then the vendors are all looking for ways to extract more information and insights out of that data so they can provide value to their customers. It’s a never-ending stream of data when it comes to security, for sure.
[04:23] Tim: One of the challenges is, then making that data a useful thing. I know this is a big problem in security, in modern security, where you can instrument and monitor so many aspects of whatever technical infrastructure you’re trying to protect. You generate so much data that is overwhelming, frankly. I mean, you can— it’s almost unbounded in the amount of data you can generate. So, I think in the work, the little work that I did with you several years ago, the big task was, can we find the parts of that data that are actually interesting to us from a perspective of trying to help the customers? Is that the challenge that you often see?
[05:11] Mike: Yeah, you know, I love that as an example, honestly, the idea there, because the initial sort of perspective, I think that Distil had as a product was to, you know, that word distill, to distill things down, to simplify, to present some really salient information to the customers. And at a certain level, the dashboarding, the reporting that we were providing was only getting so much of the insights. Maybe it was too— it was taken too far to, you know, to— in a way that presented the data at too high of a level. What we were trying to do, you know, in taking that data was, you know, I guess the basis of it, maybe since no one knows what we’re talking about, is that we were collecting all this log data with Distil acting as a proxy.
We were able to take all this log data and look at HTTP headers, look at all of this kind of wide swaths of information and really look at it kind of in the circadian rhythm of time, like on a time series chart. And the initial Distil efforts really summarized things to like a day and raw counts, and it looked really great. It was pretty, but it wasn’t very insightful in saying, hey, what’s really happening in the real world? And so working with you and looking for ways to better make data actionable in the moment, you know, kind of this is the direction I think a lot of companies are struggling with or, you know, looking to better exploit this data is to think about it as instead of points in time, things on a time series and the insights that can be gleaned from that.
I don’t know if that’s a good answer, but that was very helpful for us as a business and really launched us was those kinds of experiments and ideations.
[07:06] Tim: Which gets me back to the more meta question that we talked about as an idea for this interview, which is why from— I’m particularly interested in your perspective in interacting very closely with customers, as well as dealing with the type of products and enterprise solutions that generate enormous amounts of data. You see the ability to do useful things with that data as an invaluable tool. And maybe one that’s still under— despite all the talk about data science and analytics and all the tools we have, maybe one that’s still underused. Why is it so essential? And I guess the follow-up to that is, where are the gaps that we have that are keeping it from being even more valuable?
[08:02] Mike: Yeah. Well, I mean, I talked to a customer the other day and maybe even on you know, this sort of topic where, you know, they were saying, hey, I buy these different tools and they tell me my life is going to be easier, but instead of giving me 5 really crystal clear insights per day, they’re giving me 200 or 500 alerts that I— it’s just adding more work. And then it starts to feel like, you know, crying wolf and, you know, you know, stuff that’s maybe not even actionable or that idea of false positives and the like. And, you know, that’s one area, right?
Like when you’re thinking about developing your product and the thing that you’re trying to give to customers for operational use, that’s an area where data is super important and having a data science or data decision science team that can help think about What’s the problem we’re trying to solve? What’s the most critical bits of information? How do we deliver the exact right thing to the customer on the operational plane? But then there’s also this other side that I think it maybe is even more forgotten, and it’s the value derivation, right?
When we think about how we do a QBR with a customer or a, you know, these account management type, or maybe even during a proof of concept or proof of value project, delivering something at the end of that session that says, hey, here’s what we did, here’s what we accomplished, here’s the net value, so that those folks can go and sell that internally, right? The people we’re selling to can go sell to their internal leadership and claim budget for these types of solutions. Like, that’s another area, you know, a whole separate area, but both of them I think are super critical. And the data scientist in concert with those folks facing the customer, you know, it’s important that they’re able to work together and kind of have these as common goals.
[10:02] Tim: Yeah. So one of the things that seems to be a repeating idea that people have is the democratization of data science or democratization of analytics. Have you come across that concept?
[10:15] Mike: Well, I consider myself sort of a pretend engineer, so yes. The tooling has become so good that even someone like me, who hasn’t had a daily engineering job for quite a number of years, can jump right in. I can use tools like Tableau or even flex my muscles a tiny bit with Matplotlib, Seaborn, or whatever Python library to whip something up quickly and demonstrate a point or some attack information.
[10:52] Tim: I think when we were working together at Distil, Tableau was a product that was used by your team, if I remember correctly. And it’s an interesting product because I think a lot of people with more kind of a programming perspective on data science sort of roll their eyes at more of these point-and-click type tools like Tableau. And I’ve never actually used it, but it seems to me, just as the way spreadsheets have allowed people to do quantitative things that they probably don’t— so many people don’t have the capacity to write a program, but they can do pretty powerful things in a spreadsheet. It seems to me that they’re It really is a lot of opportunity if people can get access to data in less complicated technical ways, then there’s big opportunities for them to draw the insights and do that kind of work. Is that a fair perspective?
[12:03] Mike: Yeah, no doubt. I don’t think what you can do with simple desktop visualization tools is any replacement for a data scientist. I get that you might not consider those pro-grade tools. But at the end of the day, I can use them very quickly, almost like a proof of concept, to ideate and get that quick feedback loop with customers. Sometimes we’ve been able to take that information back to the data science and product teams and say, hey, let’s dig into this a little more.
And then that, you know, it might be just that little kindling that starts a fire and goes and, you know, takes someone in some direction. I think it goes to this point, maybe the broader idea of even why, you know, that you and I are sitting here talking. It’s that, look, the connection to the customer and what we’re trying to solve is really maybe the, you know, the most important driving force, right? I think from growing the business, growing revenue, sustaining revenue, eliminating churn, constantly driving value. Those concepts and then the enhancements that, that, that can bring to the product, um, you know, it’s, it’s a very organic thing. And, and it might start with this and it might come from the, the, you know, the other side as well.
The things you’ve provided to me, or that other guys like William or Brenton have provided from their experiments when we worked at Distil, have the same effect. Those things bring value back when we take them to customers. I love the idea of the usefulness of data: using data to drive value and drive the product in the direction it should naturally go.
[14:14] Tim: And though that actually points to one of my concerns on, from where I’ve seen data science done, Particularly in cybersecurity companies. Well, I don’t know about particularly, but where I’ve seen it in cybersecurity companies where you hire these data scientists who have graduate-level training, and often in cybersecurity, it’s hard to find someone who has cybersecurity domain knowledge. And so you kind of lock them in a room and say, here’s some data, come back with models or actionable insights, and keep them at somewhat of an arm’s length from actually interacting with customers in a way that I think is actually potentially detrimental to effective data science work. Meaning you have one or more steps between the data scientists and the customers. And there, I think you can often have a breakdown in communication in the chain between the two in knowing what is needed from the data science team that’s actually going to be valuable to the customer.
[15:38] Mike: Yeah.
[15:40] Tim: And yeah, I mean, all that to say, I think a helpful thing for companies to consider is ways for the data scientists, if only sitting in on meetings where maybe some kind of analysis or something that they’ve done is discussed with customers, maybe they don’t need to actively participate, but being able to hear what customers responses are, as well as better understanding their needs, I think would be a really helpful and important thing in allowing data science teams to deliver more valuable work.
[16:20] Mike: Honestly, if there’s a sadness or regret from my time at Distil, it’s that it took a long time before I could interact closely with you. And then when I did, it only lasted a couple of months. But when you’re considering pod architecture and putting together engineering teams, you want lots of people with different skills on cross-functional teams.
I think it really is important that, you know, especially if the product team isn’t necessarily the glue that’s able to talk between the engineering folks and the customers, that having someone that maybe sits in a seat kind of like where, you know, me or the folks on my team would sit, you know, having them stay connected to those conversations, allowing for the data engineering and data science folks to have a closer connection with the customers and really be feeling part, like they’re part of the mission and not just a tool in, you know, that someone calls upon to do something. I think that’s probably, you know, super critical, you know, especially in, you know, of, you know, the dynamics of ever-changing world of cybersecurity and the challenges that folks are facing.
And even more so now when you can’t force everybody to go to the office anymore, right? Like, you’re in the situation where we’re all going to be remote. And so tying us together in some way to give us, you know, access to you and you guys have access to us and our customers and their problems, that’s, to me, that seems like a no-brainer.
[18:18] Tim: Though I think the flip side there is the potential of data science teams getting inundated with requests that are really just curiosities. And the CEO is sitting at the breakfast table, some Harvard Business Review article, and wants, makes some seemingly innocuous requests to the data science team like, oh, how does how do we do on this metric or something? And so then they spend 4 days trying to pull down the data. Not that this has happened to me in real life. You spend 4 days trying to pull down the data and answer this question, and then by the time you get an answer back, the executive’s forgotten why they even asked the question. And so I do think it’s a tricky organizational problem of, and priority management problem of how you filter out those requests.
And you talk to anyone leading a data science team and they’ll tell you that they spend a fair bit of time telling people, no, we can’t fulfill this request, which maybe is any team. But that’s a tricky one. I don’t really know the answer necessarily how to balance other than There’s a lot of ping pong there that has to happen to find the things that are really gonna be valuable and worth the time that it takes.
[19:54] Mike: That’s why the bosses get paid the big bucks, I guess, is figuring all that stuff out because I’ll look back again to my own experience in the last decade really has been in smaller organizations. And so I I look at it and I say, hey, um, we all have some amount of ownership, maybe more than we would in other larger organizations. And so, you know, folks who want to jump in, like, let’s empower them to jump in. And maybe folks that, you know, that don’t have the bandwidth or the cycles, then they don’t. I mean, I don’t know. I think to the point of it just being delicate and a balancing act, In all organizations, that’s the case. In the ones that I’m comfortable working in, I just love the get your hands dirty and let’s figure stuff out. And yeah, sometimes we’ll throw up some stinkers, you know, bad ideas that didn’t make any sense. But sometimes, you know, you just hit one or two really good ones. Those just pull those pearls out and bam, customers are delighted. They will upgrade, they’ll talk well about you. You might save the company or earn the company millions of dollars. So those are— it’s a roll of the dice sometimes.
[21:13] Tim: Yep, no, absolutely. And my earlier point, not to take away from taking those risks sometimes, it’s not always clear necessarily what things are gonna be valuable, but I guess it’s more of, If you’re in a position where you are allowed to make requests of a team of highly paid data scientists, you need to have some kind of filter, whether it’s your own or through someone else, as to, if nothing else, recognizing that those seemingly simple requests are costly in that you’re paying for time and compute resources potentially to answer those questions. So that actually brings me back to this idea of democratized data science and tools, and I didn’t prep you for this question, but are there ways in which you think that tools that are out there now are lacking in not allowing, say, someone who’s an implementation engineer on your team Could tools be better to allow them to use the data more effectively and answer those questions themselves versus needing to rely on a data science team?
[22:34] Mike: That’s a really great question. I don’t know if the tools are necessarily the problem. I generally, if I can get access to the data, what I am finding more and more is the roadblock is you know, compliance fear, right? Like the idea of where is data going to sit? Is it going to be on someone’s laptop? What data am I going to grant you access to if you’re, you know, you’re not on the engineering team or what have you? And if, you know, so there’s all these barriers to get, you know, getting, I don’t know, I don’t want to say getting things done and compliance is a necessary part of every business. You know, products designed with compliance in mind, I guess, are, you know, they’re going to be safe and secure. But yeah, I mean, I think that’s to me is the biggest challenge. I don’t know what I don’t know.
I’m, you know, I’m just good enough at monkeying around with data to kind of figure a few things out. I, you know, honestly, if I had to say what was the tool I’m missing, I just would love to have a pure-play, you know, data scientist, or at least a really strong data analyst that was on all of my pre-sales teams, because I think that would go a long way. I mean, just that sort of concept, you know, give us that, give us data, give us someone who really knows how to use it and focuses on it, and, um, and, and let us experiment with the customer and, and give it back to product and data science to say, hey, here’s what we learned and kind of a better feedback loop. That’s maybe the bigger thing.
[24:14] Tim: I think that’s a good point for people who are interested in data science as a career. I mean, often, increasingly, people who want to do data science are particularly interested in, hey, I want to build some kind of production machine learning system, which is an important and interesting problem, but there’s a lot of opportunity and need for a data analyst who isn’t necessarily doing fancy machine learning and maybe not even fancy statistics, but just the ability to go in there and pull down data and figure out how to translate the business questions into questions you can ask from the data. And then an incredibly important thing is think about how to present that data in a way that’s going to be useful to the stakeholders and to the customers. That’s an important role and it shouldn’t be overlooked as just the data analyst versus, oh, I want to be a machine learning engineer.
[25:21] Mike: Yeah. I mean, honestly, as you’re saying it, I’m like, wow, it really just feels so important to me that you know, folks have those opportunities and can experiment and think about the business context. And, um, and even, gosh, I mean, I don’t know, maybe the question for you is, is that about— is that a valid career path for folks? Can they start as a data analyst and move to data science, or is it you’ve got to be a full mathematician and you’re not— you’re not a— no data analysts really make that move? I don’t— I mean, that’s probably a silly question, but what do you see?
[26:01] Tim: That’s a very perceptive question. And for those who don’t know, I have a blog post semi-related to this called How I Became a Data Scientist Despite Being a Math Major, where I talk about kind of my experiences to what the reality of data science is. And I partially because of that blog post pretty routinely get emails from people saying, How do I become a data scientist? And I talk to a lot of peers about this, and it’s still a very hard question to answer, partially because data science has so many different meanings that it’s kind of, how do I become a data scientist? In part is, well, what kind of data scientist do you want to be? But I do think a a really good path for a lot of people is probably not to look straight out of college for a role called data science, but maybe some kind of data analyst role or a software developer role.
And then think about if you’re interested in machine learning products or machine learning projects in particular, maybe wherever you are, start to think about how that might apply to your work. And that There’s a lot of value in academic training there, but anyone now can go read a couple of O’Reilly books and install scikit-learn and PyTorch and start tinkering around with any kind of machine learning model. So I don’t want to undervalue what training and education can offer you there, but those tools really are very accessible and available to anyone who can do some basic Python. I think, you know, not to hide here the fact that part of the reason someone might want to be a data scientist versus a data analyst is data scientists potentially make more money. And I mean, I support people trying to make more money.
[28:10] Mike: Sure.
[28:12] Tim: I don’t want to scare people away from it, but I think it’s a hard jump these days to land in the data science world where people have unreasonable expectations as well as confused expectations in companies. I think, for example, companies often overfocus on someone who has experience building XYZ machine learning model when the data scientists day to day spend a lot more hours focused on like wrangling AWS permissions and fussing with Python dependency conflicts.
[28:58] Mike: That sounds like a lot of fun.
[29:00] Tim: Which are my expertise actually. Again, not that the models and the more nuanced things aren’t important, but day to day, there are a lot of other more practical skills for people. So all that wrapping back to your question is, I think a data analyst type role is a really good place to start if you’re a young person considering this and the opportunity to move into data science or start doing more machine learning type things really are there from those roles. I mean, you can start doing that right where you are and then, you know, maybe you can get promoted internally or in a future job when you can start to talk about machine learning projects you’ve worked on. You can move on to a data science role. But again, this type of work that is Maybe more what we’re talking about with your solutions engineering projects, which is more analysis-focused, is so valuable. And if you can do that well, a company’s really going to appreciate that and I think encourage and support that.
[30:18] Mike: Yeah, I mean, no doubt about it.
[30:20] Tim: I don’t want to have the final word here. Anything more you’d like to add before we wrap up?
[30:24] Mike: You know, it’s just been really great to catch up with you, Tim. I mean, I think, you know, it’s been an awesome ride the last decade here in cybersecurity. And day to day, we’re faced with new challenges, both of us. And, you know, the intersection between our roles and the lives that we lead, I’m sure we’re gonna, you know, pass again. in ways that we don’t even know. So it’s just, it’s really exciting time to be living in a big data world and solving big data problems, even on the front lines when we’re talking to— when I’m talking to customers and, you know, trying to help build a business.
[31:06] Tim: Yeah, I’m still a little upset that you didn’t give me the opportunity to try to find you a job at BlackBerry Cylance when I was there, when you were looking for a new role, but There’s still time in the future. You still have maybe 3 or 4 more good years in you.
[31:21] Mike: Oh, thanks. Yeah, thanks. Yeah, I don’t want to age out of tech. I mean, I’m still a young man. I feel that way every day.
[31:27] Tim: Are you hiring at all at Tala Security? Would you like to do any shoutouts?
[31:32] Mike: We do have a couple of jobs open for DevOps-type roles. Those wouldn’t be reporting to me, unfortunately. Well, fortunately, let’s say. But in the future, I mean, I see the our growth trajectory and what we’re sort of thinking about for the rest of the year and looking to 2021, I, you know, there’s going to be plenty of opportunity and, uh, we’ll be looking to hire the best and brightest for sure.
[31:56] Tim: Excellent. And would you like to share where people can find you online, a website or Twitter or LinkedIn or anything?
[32:02] Mike: Yeah, my Twitter handle is @ComplicatedBull and you can find me on LinkedIn at /kfbr392.
[32:12] Tim: I like it. Thank you so much, Mike, for coming on the Into the Hopper podcast.
[32:17] Mike: It’s been a pleasure, Tim. Enjoy.