Anupam Singh - Chief Customer Officer at Clouder

(Joel Beasley at 00:00:00) Hello, my friends. Today we are talking to Anupam, the Chief Customer Officer at Cloudera, and we discussed the number one thing you should think about when going through an acquisition, the multiple dimensions necessary for predictive analytics, and why in trying times the best thing to do is invest in yourself and exercise optimism. All of this right here, right now on the Modern CTO Podcast. Here we go. This is the Modern CTO Podcast.

(Joel Beasley at 00:00:35) Hello. Hey, buddy.

(Anupam at 00:00:38) Hey, man. How are you? Good. What's going on? Good. It's—I would be traveling right now in New York, but I'm in my backyard, you know, barely insulated shed. That's where I'm at.

(Joel Beasley at 00:00:52) Dude, it's your man cave.

(Anupam at 00:00:54) It was a man cave till now. It's just become an office, so the testimonies are different now.

(Joel Beasley at 00:01:02) Well, for the next hour we're going to kick it back up to man cave. I'll send you some posters. There'll be a surprise when they arrive. We'll just man cave it up. I'll send you an ashtray for cigars and a scotch or a whiskey.

(Anupam at 00:01:23) Yeah. I have "The Last Dance" playing in the background. I don't know whether any of you are watching the ESPN show about Michael Jordan's last championship.

(Joel Beasley at 00:01:34) Oh, really? So it's a whole show, like a series or just a long episode?

(Anupam at 00:01:38) It's a series—ten episodes—and they've really lent them it. They could have done it in three episodes, but if you were ever a Jordan fan or a basketball fan, it's a very interesting thing to watch.

(Joel Beasley at 00:01:53) Yeah. I like how you mentioned the differences because I'm not a fan of basketball. I don't follow it, and I couldn't tell you about the teams and the players. But I am a fan of Michael Jordan because I've read his stories. I've watched countless things about him. I actually—I don't think I've ever heard an interview with him in it directly, like a long form interview. I've seen clips. But, man, that guy's story as far as persistence and pushing yourself and overcoming—that's an admirable story.

(Anupam at 00:02:25) Yeah. Absolutely. Absolutely. All right, how are you doing?

(Joel Beasley at 00:02:30) Good. This is it, by the way. This is the podcast. We just hang out, talk about Jordan, talk about everything. I'm doing pretty good. This morning I got to speak with John Mattone, who was one of the most—like, top three biggest leadership coaches in the world.

(Anupam at 00:02:48) Uh-huh.

(Joel Beasley at 00:02:48) Yeah. And he used to be Steve Jobs' leadership coach. Oh. Yeah. And so I got stories. I got all sorts of good leadership insight and advice from him.

(Anupam at 00:03:02) And you just downgraded yourself from business class to economy or something talking to me.

(Joel Beasley at 00:03:08) Are you kidding me? No way, dude. You're amazing. I'm actually really curious because I saw your title as Chief Customer Officer, and I was like, what is that?

(Anupam at 00:03:18) So on a good day, it is the ability to talk to CIOs. CIOs are people too. They have a lot of apprehensions, anxieties, ambitions, and whatnot, and you should be able to take technology and translate it for them. So that's a good day. On a bad day, you're just a human shield for your team. Something's not working. Customer's really unhappy, and your team is just trying to debug it, trying to figure out what's wrong, but the customer needs somebody to be on calls. I've taken 4:30 a.m. calls, especially during the—as the pandemic was picking up, a lot of the banks started monitoring for risk in the market. And that put a lot of pressure on systems, which means I had to answer 4:30 a.m. calls from customers saying things are not working. Is your team looking at this? Etcetera. So a good day, CIO whisperer. A bad day, a human shield. That's how I describe the Chief Customer Officer.

(Joel Beasley at 00:04:29) I like it. Yeah. Let's hope for more good days than bad days. Obviously, the pandemic put a stress on a lot of systems, but can you give me a quick overview of the type of service Cloudera provides?

(Anupam at 00:04:42) Yeah. So I like to tell people if you made a phone call, if you've swiped your credit card, if you bought something online since the morning, you already used Cloudera. A lot of telecommunication providers across the world, whether it is Philippines, India, UK, or America, are managing their networks through analytic dashboards built on Cloudera. So you call into your Internet service provider and say, "Hey, my Internet's running slow." The first thing they do is check whether your ZIP code has any Internet problems. That dashboard is highly likely working on a big data system. Or if you see a suspicious transaction on your credit card and again you call your bank or your credit card provider, they pull up a dashboard of, "Hey, you know, what are Joel's last three transactions? Do they match his pattern? Are they in a country that you're already in?" All of that, most likely, that data was prepared by a Cloudera system.

(Joel Beasley at 00:05:54) That's pretty cool.

(Anupam at 00:05:55) Yeah. Yeah. Yeah. But it's all invisible technology, so it's not as if you're talking to Cloudera at all. My mom keeps asking, you know, "Why don't you build something like WhatsApp?" And I say, "But how will the WhatsApp analytics run on top of Hadoop?" And I lose her attention completely the moment I mention one of these technologies.

(Joel Beasley at 00:06:16) Come on. Your mom doesn't know about Hadoop? That's my problem. Oh, man. That's funny. Now were you one of the founders of this company? I saw you've had some entrepreneurial background.

(Anupam at 00:06:28) Yeah. So I've founded two companies in the big data space. Wasn't the founder of Cloudera. The founders of Cloudera bought my company in 2015, so five years ago. And, as the story goes, most entrepreneurs want to leave within a couple of years. But somehow at Cloudera, there's a bunch of us who have all been founders of their own companies, but have come in through acquisitions and have stayed back. Some for four years, some for five years. So I'm one of those.

(Joel Beasley at 00:07:04) That's pretty cool. I could—to imagine—yeah. I'm a founder of a company and we've grown. And to imagine what it would be like if I were to sell this venture and then go work with a group of other people who've gone through the, you know, staring into the abyss and eating glass that is founding your own company. It just must be an unbelievable energy just working with high-performing people. Is that what it's like?

(Anupam at 00:07:34) So one is gratitude, all right? Because when you're an entrepreneur, you feel very alone. You feel that the burden of every decision is on your shoulders. What's worst is when you have a bunch of people who have walked on shredded glass or eaten shredded glass, they are able to take some of those burdens. They can take the 4:30 a.m. call, or if you call them at 5 a.m. in the morning, they don't completely melt down. So I have a lot of founder colleagues who have seen tough times, and even apropos today's times, it's very good to have them on your team. You know, you can sort of pass the ball and get a breather, get a timeout. That's the beauty of working with entrepreneurs.

(Joel Beasley at 00:08:23) So can you tell me a little bit about what the rise of the enterprise intelligence platform is?

(Anupam at 00:08:31) So the enterprise intelligence platform—the way I like to explain these days is, in the context of today's times, everybody has become data scientists. You know? People are looking at dashboards. I look at Santa Clara County dashboards, for example. Nobody understands what bending the curve is, but everybody's talking about bending the curve. What's happening is most of us are starting to look at data differently. Most of us want to be more intelligent in our decisions, whether it is buying stocks, whether it's buying cars. You'll never go to a car dealership without looking up the last few sales in the area, how much they were. So pull that back—every enterprise, every bank, every telecommunication company, health care company needs to provide you that. For example, you wouldn't go to your health care company if they didn't have an app and if they didn't show the results of your last few tests on the app. You'd regard that as primitive. But we forget that even ten years ago, you wouldn't be able to do that. That's intelligence. Right? Intelligence is a bank telling you about a transaction that's suspicious in real time. Behind that, there's a lot of analysis of data. So imagine there's trillions of credit card swipes that are going on. How did they find that this particular transaction needs Joel's attention? That's intelligence. So that's what enterprise—every bank, every telecommunication, every health care, every government, every grocery shop for all you need—needs all that intelligence because customers are demanding it. That make sense?

(Joel Beasley at 00:10:22) Yeah. It sounds like you've got a lot of fun. You have a very fun job getting to build these types of products.

(Anupam at 00:10:31) Yep. Yep. Yep. Yeah. Yeah. And see it having effect but with an invisible hand, right? So that—if you're an infrastructure software guy like I am, knowing that when you pick up your phone or when you interact with your pharmacy, a lot of it is being powered by our systems is a lot of fun. And by the way, I started in this Hadoop big data space when it was fourteen or fifteen people meeting in a small room talking about an open source project. The fact that it is now billions of dollars of revenue, etcetera—none of us thought that it's going to happen. It was more of a passion project for a lot of us.

(Joel Beasley at 00:11:20) Oh, so you were part of one of the original Hadoop teams?

(Anupam at 00:11:25) I was one of the original groupies of the Hadoop teams. That's how I like to think about it, right?

(Joel Beasley at 00:11:31) I like it.

(Anupam at 00:11:32) There were much smarter people than me—Doug Cutting, Arun Murthy—who were articulating the future. And I was, you know, I was the groupie saying, "Yeah. That sounds really interesting. How can I help?" And that's still my role: helping customers understand this complicated technology. But to see that from fourteen, fifteen people to where it's at is still surreal for me every day.

(Joel Beasley at 00:11:59) Yeah. It's like a household name in data science.

(Anupam at 00:12:02) Yeah. It's just weird, honestly. You know?

(Joel Beasley at 00:12:05) And that groupie aspect of—or being a supporter of something early—is critical to the success of it as well. I mean, it's very important to have people pushing that forward. It helps with the energy of the core team, right? And it's just—it's a very valid part. Now I want to get a little nerdy with you because I've gotten some questions from people. But the question I'm getting a lot is, "Joel, we're getting into data science. We've hired some data science people. We're looking at these metrics." But the broad spectrum question from all the different conversations is people are asking me, how much data is enough data to make a decision on?

(Anupam at 00:12:51) How much data is enough data? Couple of points of view on that. Number one is when you're prototyping, you don't need a lot of data. And I'm going to get uber geeky on you. I'll double down on the geek, right? You need a lot of metadata, right? Is Joel a name, or is it a product? Right? Michael Jordan is a name. Air Jordan is a shoe. These are two very different things. So just dumping data and being happy that it is a lot of data doesn't make any sense unless you have metadata. So that's the first big thing that people miss, right?

(Joel Beasley at 00:13:36) So can I bring it down with an example, like a closer example? Like, something we're all familiar with—email subject lines. Right?

(Anupam at 00:13:45) Yeah.

(Joel Beasley at 00:13:46) So let's say I'm sending emails and I'm testing two different subject lines. How many emails do I need to send before I can tell which one—if I'm going to run a test, how do I know how many emails I need to send before I find a winner?

(Anupam at 00:14:01) Yeah. So for that, the biggest thing that you need is the other parts of—we call it dimensionality of data. Who are you sending it to is important. When are you sending it? How big is it? Right? Did they open it on the phone or did they open it on an email client? Right? And did you add pictures or not? So already you added five dimensions to the data. A lot of times, people think data science is about looking at the subject line and trying to find Joel's tone and whether it worked or not. Right? That's picking up just one line and trying to be this amazing—geeky natural language parsing is very fashionable right now, right? And try to figure out whether Joel's tone works or not. No. It's all the other context around the data. When, how, who opened it—you know, where did they open it? Did they open it on a jog or did they open it, you know, while they were watching TV or were they at work? There's so much context that you can add to the data. So adding more context is as important as just sheer raw computing power with a lot of data. That make sense?

(Joel Beasley at 00:15:17) Yeah. It does. 100%. And then in my search for trying to figure out how much data is enough data, I did come across one concept about statistically valid sample sizes, but by no means am I an expert. Have you come across this?

(Anupam at 00:15:35) Yeah. Statistically valid sample sizes, I mean, you know, it depends on what you're trying to do with the data. If you're just counting—and counting is very unglamorous, by the way. A lot of data scientists think that they have to predict. There's a fine line between counting and prediction. If you are trying to predict, then you're going to have a very different point of view. You're going to use neural networks. You're going to use machine learning. But if you're just counting, then you need to be fast and you need to count a lot of different things. Does that make sense? So counting doesn't really like sampling that much. Whereas predicting likes to sample a lot of Joels and Katies and Jakes and Anupams to get to predicting Joel's behavior using other people's behavior.

(Joel Beasley at 00:16:34) That's an interesting line. I did not have that differentiation in my mind, counting data versus prediction of data. What do you do mostly at Cloudera? Is it mostly prediction?

(Anupam at 00:16:46) So I personally came up through the counting side of things, which is I have hundreds of terabytes, and I can look at everything that Joel did on the Internet, and I can count. Okay? My smarter colleagues, who went to school later than me, actually are from the field of what is now called data science and a bad way to say it is artificial intelligence. Right? So a lot of our systems are now pivoting towards artificial intelligence and machine learning. So while Cloudera's original business mostly is still counting a large, large, large amount of data, the growth we are seeing is in things like machine learning.

(Joel Beasley at 00:17:32) So that's the big area that's growing right now, machine learning?

(Anupam at 00:17:36) Yeah. Yeah. So imagine, for example, you know, an airline has to maintain kiosks. Right? And every time you send somebody over to fix a kiosk, it's going to take effort. But what if the kiosk could tell you that I'm about to fail in the next twelve hours? You would be able to send somebody over. So that's a prediction example where a kiosk is telling you that they're going to fail. But the best story I have on this—so Philippines is actually a country with thousands of islands. Correct?

(Joel Beasley at 00:18:13) Mhmm.

(Anupam at 00:18:14) Okay? But it's also one of the most advanced telecommunications countries in the world. So everybody has a cell phone.

(Joel Beasley at 00:18:22) Oh, because of the landlines. Yeah. Yeah.

(Anupam at 00:18:25) Where are you going to put landlines? Got it. Thank you. Right?

(Anupam at 00:18:28) So you don't have landlines, so you're putting towers in small islands. Anytime a tower fails and I have to send Joel to fix it, it takes $50,000 because Joel is going to take a helicopter.

(Joel Beasley at 00:18:43) Nice. Or a boat. A nice helicopter or a yacht. I'll take a yacht.

(Anupam at 00:18:46) So you'll take a yacht. So that's $75,000 now. If the tower could predict that I'm about to fail, right, then I'd make sure the guy goes to the island, but goes there to a preset time.

(Joel Beasley at 00:19:02) You're going to send me on coach. You're going to send me on coach early. Oh, no.

(Anupam at 00:19:08) So no helicopter, no yacht. I'm going to send you, right, on an economy minus ticket and see if I can get—

(Joel Beasley at 00:19:16) I'm in the cargo hold.

(Anupam at 00:19:18) Yes. Exactly. So that's a live example. This system is saving hundreds of thousands of dollars, right, in Philippines right now. The customer is Globe Telecom, and they predict that this cell phone tower is about to fail.

(Anupam at 00:19:37) It's pretty amazing, actually.

(Joel Beasley at 00:19:39) And then, obviously, there's some code that, like, what component's going to fail. It gives them some diagnostic. You know, here's a question. How do you guys—I'm going to guess at how you do it and you tell me how far off I am. Do you take past data from a bunch of failed towers and create a machine learning model around that so it learns how these towers are failing off of this data and then apply that model, or do you do it differently?

(Anupam at 00:20:05) That is one technique, all right? But firstly, there's very unglamorous parts of it. Imagine how much time it takes for that data to show up in a central place where you can analyze it.

(Joel Beasley at 00:20:17) I don't know. How much time does it take? Right?

(Anupam at 00:20:19) So there used to be this type of work would take almost a month for the data to show up, to be clean, to be ready for analysis.

(Joel Beasley at 00:20:29) Oh, wow.

(Anupam at 00:20:30) Okay? Today, it happens in minutes. So firstly, you want more real-time data. Because otherwise, anything that you predict is going to be old news. Right?

(Anupam at 00:20:42) So that's one part. The second part is that it's not just the failing towers that you want to see. What you want to see is when a tower is successful, what does the healthy cell phone tower look like? Maybe it's not heating up. Maybe it has a temperature range.

(Anupam at 00:20:59) So you combine a well-working tower with something that went badly, and now you say, what is the difference between these two? And then anytime you see this difference, anytime you see temperature drop, maybe weather, maybe rain, maybe humidity. Right? When you know the good range, then you can identify the one that is out of range. You may not know why, because the worst thing about predicting is you also try to predict why Joel sent an email at 1:30 a.m. in the morning.

(Anupam at 00:21:31) I don't know. So instead of explaining the why, you just see that it is out of range. This tower is behaving out of range. So you proactively land it before it catches fire. Being dramatic here, but, you know, that's the kind of stuff models are supposed to achieve.

(Joel Beasley at 00:21:50) That's right. Because then that's going to affect the cost of the business, not only sending the person out there, but loss of business from the interruption or stress on customer service from the interruption. There's a whole lot of ways—

(Anupam at 00:22:04) Exactly. Impact. Yeah. We have seen numbers from another telecommunication company that if you fix something within 15 to 20 minutes, your call volume is much less. But if it goes on for half an hour, then your call center gets overwhelmed, and now you're trying to fix the call center because too many people are calling up with the same problem.

(Anupam at 00:22:26) So you have that golden period in which if you fix it, lazy people like me will not call my telephone company.

(Joel Beasley at 00:22:35) Yeah. I guess they're—I don't know how they're doing in the Philippines, though. They go to a payphone.

(Anupam at 00:22:42) It's very interesting. What they do is they move to the other provider.

(Joel Beasley at 00:22:46) Oh, so they just—so—

(Anupam at 00:22:48) In Asia, you know, across all of Asia, people carry two SIM cards.

(Joel Beasley at 00:22:55) Oh, really?

(Anupam at 00:22:57) So if I'm not happy with one, I'm just going to switch over to the other SIM card. It's your loss. I'll move from Joel's company to Katie's company, and Joel just lost revenue. And then you come back online, I'll move from Katie's company back to Joel's company. It's fine.

(Anupam at 00:23:12) Because it's literally two SIM cards in the same phone. You're not carrying two phones.

(Joel Beasley at 00:23:17) Yeah. I've seen that. I was curious about that because I've seen the phones where you can have two SIM cards in them, and I was curious about that.

(Anupam at 00:23:24) Yeah. Yeah. Yeah. So loss of revenue is very, very real there because also a lot of the subscribers are not the kind we are. Right?

(Anupam at 00:23:34) We most likely, you have an annual subscription to a phone provider. Right?

(Joel Beasley at 00:23:38) Yeah.

(Anupam at 00:23:39) To your telco. There, they have SIM cards that need recharging every six or ten days.

(Joel Beasley at 00:23:47) Got it. That's interesting. Oh, yeah. There's a lot of towers out there too. I mean, there's towers everywhere. There's only becoming more of them.

(Anupam at 00:23:54) Oh, yeah. There's—as you pointed out, the cost of putting a landline is impossible in many cases. It's not just high. It's literally how many fiber optic cables am I going to put across 3,000 islands? It might as well put two towers and be done with it.

(Anupam at 00:24:14) So, yeah, there's tons of that. In India, Reliance has gone from around 100 million customers to 400 million people. That's the population of the United States, which has gone online on cell phones through towers in the last two years, I think.

(Joel Beasley at 00:24:38) Yeah. I was talking with the CTO of Google the other day, and they're talking about how, I think, in Africa, they put these—I think it was hot air balloons that would broadcast internet down so that they could come online because it's cheaper to do it from the air. I know Facebook had a program where they were doing gliders, and apparently it's cheaper to do that than it is to actually run the infrastructure on the ground.

(Anupam at 00:25:01) Yeah. Our customers call it rolling a truck. If you can avoid rolling a truck—

(Joel Beasley at 00:25:06) Mm-hmm.

(Anupam at 00:25:06) You're saving money. Yeah, of course, in the Philippines, you're rolling a helicopter, which is even worse.

(Joel Beasley at 00:25:12) Well, it depends. Depends on who you are.

(Anupam at 00:25:14) On the helicopter.

(Joel Beasley at 00:25:17) 1:30 a.m. helicopter. I was just trying to impress a girl.

(Anupam at 00:25:23) And costing a lot of money.

(Joel Beasley at 00:25:26) I know.

(Anupam at 00:25:27) Costing a lot of money to the company.

(Joel Beasley at 00:25:29) Oh, by girl, I mean, I was trying to impress my daughter because she wouldn't sleep, and so I was trying to take her on a helicopter ride. By the way, that's how we get our kids to sleep as we put them in the car and drive them around. Puts them right to sleep.

(Anupam at 00:25:39) Yeah. Oh, my—yeah.

(Joel Beasley at 00:25:41) Let's talk about cloud computing. You have a lot of experience there. I am curious, like, what does the next ten years look like for cloud computing?

(Anupam at 00:25:51) I think the biggest thing is going to be privacy. People are going to start—my 68-year-old mom asked me the other day, I'm downloading this app which says the government is going to use this for COVID tracking. Okay? No problem. How much of my data is going to be saved by the government for other users?

(Anupam at 00:26:14) She's like, wow. So you're thinking about privacy. I mean, she didn't use the word privacy, but the fact that she was thinking through how much of her data will be used, and then she had a lot of questions. Can they look at my WhatsApp? Can they look at my phone contacts?

(Anupam at 00:26:31) Right? And I was just amazed at her literacy on data governance. That's what we call it. Right? So I think cloud computing is going to have an interesting reckoning around security, governance, privacy of data.

(Joel Beasley at 00:26:48) Did you see that the Senate, they voted to let the FBI access browser history without a warrant?

(Anupam at 00:26:56) Yeah. So all of that is going to be very, very fascinating how that unfolds, correct? Right? Whether it's browsing history, whether it's your phone history.

(Anupam at 00:27:07) There's a lot of data that we're all spewing. So all of this data is going into the cloud. Fair enough. Right? But one, brands will have to build that trust.

(Anupam at 00:27:19) But without naming names, there are certain brands I'm sure you trust with your data.

(Joel Beasley at 00:27:24) Yeah. But the one brand that I don't trust is the only brand that I need to access the internet, and that is whatever brand is providing me connectivity. Right? Like, I've been an engineer for 17 years. You only need the—your ISP. Once you have the law in the Senate and they have the connection to the ISP, they can just suck that data out right away.

(Joel Beasley at 00:27:46) They don't need to get Chrome's permission to get my browser history. What they're really just talking about is my web traffic history. And so I think we're definitely going to see a lot of rise of people becoming intelligent and aware of how to use VPN-type systems or browsers that allow for privacy beyond your ISP.

(Anupam at 00:28:10) Yeah. And so that's a heavy area of investment for us as a company as well as where consumers' mindsets are going. Right? In the current times, people are willing to share a lot of data with the local government, where you went, who did you meet, tracking, et cetera. But they are going to start asking the question as we recover is how much of my data is going to the government?

(Anupam at 00:28:39) And what is the policy behind it? Can my insurance company access it? All right? Can my clothing company access it? So I think that's a big area.

(Anupam at 00:28:54) The other one is going to be in cloud computing is going to be choice. Everything starts with one big provider. You know, whether it is messaging, whether it is phone, it starts with one big provider. But steadily, people, enterprises, banks, et cetera will start asking, but I think I need two providers or three. Right?

(Anupam at 00:29:15) So cloud computing will have more and more players, whether it is Google, whether it is Microsoft. They're going to start challenging, if you will, the lead that Amazon Web Services has today. So that's going to be interesting for a lot of us. And the next exciting thing is how much compute power you and I will have as individuals. The infinite compute power.

(Anupam at 00:29:44) But here's the problem. With infinite compute power comes responsibility, and that is the cost. Right? That is the cost. Bills will go up for cloud computing.

(Joel Beasley at 00:29:58) How much money did you spend looking at cat memes, you know, Grandma?

(Anupam at 00:30:01) Yeah. Yeah. Yeah.

(Joel Beasley at 00:30:02) Grandma.

(Anupam at 00:30:05) And if you're a bank, somebody decided to run a model and then got that cost to—I've had a customer where this person runs a report. That report cost them $625,000. One report.

(Joel Beasley at 00:30:21) I hope it makes them money. Like, I hope the value of that—I hope it's a $2 million valuable report.

(Anupam at 00:30:29) Yep. I think cost attribution, cost projection, et cetera, will be extremely important in cloud computing. Right? What is the cost?

(Joel Beasley at 00:30:37) Why am I so resistant? I think it's because I'm American and freedom is so ingrained in our culture, but I'm just incredibly resistant to this idea of people collecting and amassing the data on us. Like, I just don't like it because I feel like they're going to make predictions of—I'm going to stop becoming Joel, and I start becoming an inhuman piece of a larger pool times a thousand. Right? Because I'm constantly in all these other groupings based on, oh, maybe they've seen my camera and they know I have a beard now. Right?

(Joel Beasley at 00:31:18) Or maybe they can tell I'm wearing a Jeep shirt, so I might, you know, own a Jeep, or they've already seen my payment history or my search history, and they see the fact that I pay my Jeep bill online. Yeah. And they start profiling me as an individual. And the problem that I have with that is that at the moment, at the time of profiling, they are considering me a static individual, and I'm a dynamic person. I will be wearing maybe a Tesla shirt next year.

(Joel Beasley at 00:31:49) You know? Yeah. I don't like being boxed in.

(Anupam at 00:31:53) But as good Americans, we also like free stuff. Let's remember that. Right? So a lot of times you're also using a lot of free stuff. Chrome is free.

(Anupam at 00:32:02) You know, WhatsApp is free. There's a lot of free software around. That's number one. Number two, it does benefit you also. All right?

(Anupam at 00:32:12) When it gives you a recommendation, like, typing ahead has made me a better writer. You know what I mean? When you get an email and you can immediately respond because Gmail filled out a response for you. Sometimes it's a little inauthentic, but at least I responded to your email. I said, "Good job," or "Thanks."

(Anupam at 00:32:35) It's a quick thing. So there's a lot of goodness to all this prediction. There's a lot of goodness to understanding what Joel wants to do next and giving you that thing. But your point is valid. It might categorize us into behavior, and we almost can't change the behavior because the system won't let us do that.

(Anupam at 00:32:58) That's going to be a very interesting problem for AI going forward.

(Joel Beasley at 00:33:02) Yeah. It's almost like if you're in a relationship and you're always predicting the person's going to do something else, eventually, they will form into that just out of your sheer relentlessness. You know, that—or that's a possibility it can happen if it—you know, what is the phrase, you see yourself through the eyes of other people? Like—

(Anupam at 00:33:24) Yeah.

(Joel Beasley at 00:33:24) That's a common one. So what happens when these algorithms and predictions become the eyes of other people in which we're seen? It's an interesting conversation, and it definitely isn't solved in a blanket statement. Right? These are things that will require lots of thought and lots of nuance. For example, you know, if I'm looking at data privacy, I should be able to toggle something that says, here's all of my data profiles, maybe like a location data profile, a search history data profile, all of these profiles.

(Joel Beasley at 00:33:57) And then here are the sources that are allowed to consume them. Yeah. I'll give it to my local government. No. I won't give it to the federal government.

(Joel Beasley at 00:34:05) And then I could choose. Yeah. You know, because what I would do is I would take my purchase history, and I would let that be shared with retailers and, you know, because why not? Like, all the things I'm purchasing, all it's going to do is going to help me purchase better things. And they're not deciding for me.

(Joel Beasley at 00:34:23) They're just offering me up things I might be interested. But when you get into, like, my insurance company rating me and charging me more money—

(Anupam at 00:34:31) Exactly.

(Joel Beasley at 00:34:31) As opposed to someone else based off of habits that I have unwillingly shared with them, that is, for me as subjectively, I guess, an individual, that's a line. Like, I should have—and I'm not saying the connection shouldn't exist. What I'm saying is the choice should exist.

(Anupam at 00:34:49) I think the choice will happen. Okay? The who can you share it with? GDPR is already doing that. CCPA.

(Anupam at 00:34:57) There's a lot of regulation work. That's one part of it though. The other part that was very interesting that you talked about is what conclusions is the data reaching about Joel and are they valid? Part one is I don't want to share my data. Joel doesn't want to share his data with Katie.

(Anupam at 00:35:16) It's fine. You know, that's done. Right? But on the other side is what was the data used for? What conclusions were reached?

(Anupam at 00:35:25) And did these conclusions have bias? So one of the things that—very interesting project that we're working on is that you have to have detection of how a machine learning model reached the conclusion. So to give you an example, if it used gender to reach a conclusion, then does that model have gender bias?

(Joel Beasley at 00:35:52) Well, I think bias is inherent inside of everything. It's the thing that keeps us alive. It's the thing that allows us to eat the food that won't kill us when we are foraging. I mean, I don't think we'll get to a bias-free. I mean, I think we would die without bias. It's just we focus as a society, as like a global consciousness, on which biases we are socially discussing today.

(Anupam at 00:36:17) Yeah. And it's important to track those biases. Right? The way you are tracking who is using my data, that is one part of it. But what is important, what are they doing with my data? Is it used to introduce bias into a model? Right? And then track it. And tracking is what we do. As a company, tracking is what I do as an engineer. We don't come up and say you should not use gender or age, because in many cases, gender and age is a good predictor of what you want to buy next. But if you use body weight to refuse health care, that's a problem. Right? So our job is to provide the information saying this model used this attribute to reach this conclusion about Joel.

(Joel Beasley at 00:37:06) I like that. I like the transparency. Yeah.

(Anupam at 00:37:09) Track that transparently and then tell the consumer. I think where the future will land is Joel will want to know why he was charged 20% extra from last month. And the companies that have trust in their brand will then explain to you why we charge you 20% more.

(Joel Beasley at 00:37:33) I love it. And take it a step further. You know, they explain why and then offer a corrective path for you to alter the behavior. You know, we charge you because you have a traffic ticket from three years ago, which by the way, before all the modern data scientist stuff, they were already hooked into the DMV and rating you based off of traffic tickets. Right?

(Anupam at 00:37:54) Yep.

(Joel Beasley at 00:37:54) So that's been around as long as computing, maybe even before computing has been around. But yeah, you know, that transparency of them telling us and then us having a path to correct it, and then us having a choice on top of it. Like, there should be a choice where I say I'm not giving you my driving record, and I subsequently will pay a higher price.

(Anupam at 00:38:14) Exactly. Yeah. That's the input choice.

(Joel Beasley at 00:38:17) Yes. We can—I will gladly, like, for us, a big difference between my wife and I, here's an interesting thing. She will never pay for any ad-free service. She just does the ads, and it drives me bonkers because every single service which I could pay ad-free for, I go ad-free because I value my attention and my ability to get into something, and I don't want to get yanked out of it with an ad. And so for me, Pandora price doesn't interrupt my workout. Right? I pay for my Pandora, my workout doesn't get interrupted. I pay for my YouTube, my watching experience—and I'm watching Joe Rogan podcast—doesn't get interrupted. Like, nothing gets interrupted, and that's why I'm paying. And I would gladly pay an extra $25 a month to my ISP called, like, a don't send my data to the FBI fee. Like, wouldn't you? Do you want the—I mean, like, I would pay $25 for them not to give my data to the FBI.

(Anupam at 00:39:21) That's interesting if you crystal ball this. Right? Will we come to a point where people will say I would rather pay for this software the way we pay Netflix for no ad interruptions? Right? I could pay my messaging provider saying, hey, take a monthly subscription fee, but please don't use my data for anything else because that's what I'm paying you for. Right? I think there's something there in that idea. I think that reckoning is coming because people are becoming more and more freaked out when they're being—as you said, when they're being categorized into a bucket. And as human beings, we like to evolve, so we don't want to be categorized into a bucket. All right? You might shave. You might buy a different car, like you said. So in that world, disallowing usage of my data could be a very interesting thing in the coming years.

(Joel Beasley at 00:40:21) Yeah. Yeah. Because I'm scanning Instagram. Yes. Another dude on a video trying to tell me how I'm going to make a million dollars. Like, I'm not—I would rather pay Instagram $5 a month and never see a sponsored post, or $25 a month and never see a sponsored post. And then here's the other thing. I think that they could make money this way because not only would I pay for not having ads and protection of my data because I value the service—you know, I use it, I like it—I will pay for things I use and like. Don't you? I mean, we do. I pay for Zoom. I love it. And so for the moment I'm ready to make a purchase, if Instagram had a tab of, like, you know, Instagram vendor deals or whatever, and I could say, hey, I'm looking to purchase a, you know, leadership training system. What do you got? And then it's going to show me, you know, a bunch of ads, and I can sit there and scroll through the ads. Okay. We got this company's advertising this way, you know, and I can just, you know, click through the companies, have some sort of like directory, and I could shop. That would be—that makes sense to me. But what they're doing, the model today is let's find any space of attention and cram an ad in there. And I don't like it.

(Anupam at 00:41:38) But I don't want to—I want to be careful about just thinking of all ads being bad because, you know, especially I've always been impressed with Google on its ability to give me the right information at the right time at the right place. I think Amazon does a good job too. So on the other end of the spectrum is if I can explain to Joel why you got that recommendation, and you'd be able to interact with it and say you were wrong. This ad was inappropriate. I've seen this before. I've already bought the item. If somebody could make that simple and then loop it back into their recommendation system, that could work too. So yes, there is a chance that a bunch of free software becomes subscription software. You pay for the subscription. Just an option. Yeah. Yeah. That is an option. But the other option where you can actually inject yourself into that decision-making process, saying, hey, your recommendation is off. I think that'll help too. And that's what the learning and machine learning is about. Learning from your feedback would be a very important loop.

(Joel Beasley at 00:42:52) Have you seen that on Instagram? They actually have that. I can, like, click don't show me this ad and they ask why. And it's like I purchased already or I'm not interested in this item. Or so they do have a little bit of that.

(Anupam at 00:43:04) They have a little bit of that, but, model—going geeky for a minute. Right? Model explanation is a very, very hot topic right now in data science because these are massive networks that are trying to build your recommendation. The amount of machine learning that happens before Netflix recommends your movie is amazing. Right? So to get you that explanation is going to be harder. So right now, the explanations are simple. Ad is inappropriate. I've already seen the ad. I've already bought the product. Right? Okay. But I think the explanations are going to get richer, and they'll be able to tell you that I showed it to you because you live in a big city with a lot of public transportation. Like, that level of explanation is much more personalized than just saying, hey, it's an inappropriate ad.

(Joel Beasley at 00:43:54) That's true. If—here's a question for you. If one of your favorite services did have, like, a tab that was, like, suggested suggestions, you know, like different products. Do you think you would go browse through that just to see? Like, if I had an interface that said here's some suggested products for you and here's why based off of all the data that exists out there from all the third-party providers that collect your data and aggregate it. I think that would be pretty—I would look through that from time to time to see, like, the new recommendations and what they think I would want because, you know—I don't know. What about you?

(Anupam at 00:44:30) I think it's already happened. It is that we are not spatially aware of it, but every time you browse a product on any e-commerce site, it also gives you suggestions. So that's happening already.

(Joel Beasley at 00:44:43) Yeah. Jake just messaged on the chat, and he said they have some of that in Facebook. But they don't have it, like—it would be cool if they had it, like, all in one area because then I'd be like, this would be—there are times when I—I mean, people as humans, we're used to this concept of going shopping. Right? But right now, all of our shopping interactions are meshed into this feed whether we want them or not. It's like, let me do my feed when I'm doing my feed, and let me shop when I want to shop, you know? And they'll get paid either way. I still shop as much as the next person. I'm not more or less like—like I would spend time in both places, and I would have a higher level of intent.

(Anupam at 00:45:24) What we have seen in behavioral patterns is people say that, but really, a lot of the recommendations are better than organically. Meaning, while you're doing one thing, it suggests another thing, and you're okay with it. Right? I know. That's the beauty of being in technology. Right? There's some very silly ideas. I thought virtualization was silly. I thought Linux will never pick up. Right? All kinds of interesting things. Amazon will only sell books. And then you'll see a lot of other things happen. So I think the biggest thing is going to be explaining to you why I'm asking you to look at this product, whether I ask for you to look at it in line or whether I ask you to go somewhere else like a separate tab.

(Joel Beasley at 00:46:19) I'm following. That's like—sorry. We got off topic craziness. This is—I like talking to you. This is a lot of fun. So we were going with, like, the 10 years thing, and so we think more explanation and more understanding of these models will appear within our day-to-day lives.

(Anupam at 00:46:36) Yeah. Yeah. Absolutely. Nice.

(Joel Beasley at 00:46:38) And then I guess that was a broad question. Cloud computing. Right? Like, it's a pretty broad area. How do you even, like, define cloud computing?

(Anupam at 00:46:53) Cloud computing is essentially when you're not racking and stacking computers and where you are paying as you go. Right? These are two important things. And the third one is a softer one, which is that you're also not really buying the software. You're not installing even the software. To think of a software project, the first is you get some servers, then you deploy the software, and then you buy a newer version of that software, simplifying the entire life cycle. Cloud computing eliminates all of that. You never buy a server. Right? You don't really install the software, and you pay only for the times that you use it. Makes sense?

(Joel Beasley at 00:47:45) Yeah. That's a great explanation. So you got acquired from this company and now you're in there. What is one of the things that you learned through that acquisition process?

(Anupam at 00:47:57) Big thing is taking care of the team as it gets acquired. All right? So for a lot of first-time entrepreneurs, acquisition itself is act three, climax, championship won, whatever. Right? It's very—it's sort of the ending of the story. I've learned, having been through two acquisitions and having acquired five companies for Cloudera, that it is just the beginning. It's act one. It is then landing the team, making sure that they are finding productive uses of their time, and their careers are being taken care of. Some companies do it well. We like to think we do it well too.

(Joel Beasley at 00:48:41) Do you like this process in your professional life of getting to meet these companies, go through the negotiation process, and acquire them, and onboard them? Is that fun for you?

(Anupam at 00:48:52) Yeah. It's very fascinating to—if you see an entrepreneur at the first time they have an idea. So now I have a list of 10 or 15 almost-billion-dollar companies. Some of them we just discussed in this backyard as, hey, I think I have an idea. It's not even an idea. They think they have an idea. Right? And then you see it on a billboard, and you say, wow. I didn't realize video conferencing could be disrupted. I didn't realize databases could be disrupted. So the joy of seeing an entrepreneur starting with a small kernel of an idea and then being able to take their idea and giving it a bigger stage and then a bigger stage, it's a very satisfying feeling. And some of these people I grew up with in the industry, we used to dream of building companies, building products that people used. To see them now running their own companies or being able to bring them in and keep them at Cloudera and still have that entrepreneurial hustle, it's just—yeah. Makes my day. I love it.

(Joel Beasley at 00:50:04) Well, as we start to wrap up here, is there any topics that we didn't cover that you want to talk about?

(Anupam at 00:50:11) I think a lot of people are struggling to figure out how their careers will be in the next few months, years. There's a lot of bad news, and I keep telling people, having been through two different recessions—2000, 2001 was pretty dark. 2009, 2010 was pretty difficult times for people—is continue to invest in yourself, exercise optimism. And the word exercise is very important here. You have to make an effort. Don't check news every three minutes. Listen to a podcast like your podcast and think about growth in your career because things will recover. And finding your optimism right now is difficult, but it's very, very important for people's careers.

(Joel Beasley at 00:51:04) I love it, and I agree. Because whenever I get down, I realize that the way life works, it's like an up-and-down thing. So when I'm down, I'm like, this is actually not as bad of a thing because the moment you realize that you're down, you're like, well, it's only up from here.

(Anupam at 00:51:22) Yeah. And for technical people, Google was founded in the throes of a recession. Right? Zoom, Okta, Twilio, Cloudera—all of us were formed during the last recession. So right now, one of your friends, one of your old managers, one of your bosses, somebody is building an idea either inside the company or outside the company. Go gravitate towards them and stop checking the news every three minutes. Go ask your mentors what are they thinking about. Is there a novel innovation that they've been thinking about working on within the company? And latch yourself onto that project because recovery will happen, and people who invest in themselves right now are going to be the leaders of tomorrow.

(Joel Beasley at 00:52:05) Yes. Preach. Yes. Right? When this whole thing happened, I was like, I want to be the guy that when this thing is over and everyone's back, that I look healthier, that I'm happier, that, you know, I have worked on my fitness, that I've worked on my mindset, and that I come out of this thing with a net positive. That doesn't mean I don't have the flexibility to have my doubts and my fears and go through my human process, but it means that I will come out of this better than I went into it.

(Anupam at 00:52:37) Yep. Exercise optimism.

(Joel Beasley at 00:52:40) I love that. You know, I don't think I've ever heard it explained like that. I think you articulated it beautifully because people inherently feel like, oh, I'm just not—it just doesn't happen for me. It's like a jacked, ripped person. It's like, no, this just doesn't just happen for them. They exercise it. They make it happen, and they consistently execute on it and exercise it over long periods of time, and that's how they become somebody that they're proud of.

(Anupam at 00:53:11) Yep. Yep. Yep. Absolutely. I stole it from—my full disclosure, I stole it from my son's middle school principal, but I like it, so I wrote it down.

(Joel Beasley at 00:53:20) That makes it even better because the advice is so good. It is. It really is good. Oh, man. This is great. I'm so excited you came on and you hung out with me and we got to talk about this stuff. It was—I was not expecting the conversation to go with privacy, but I've been thinking about it a lot. I'm glad we got to talk about it.

(Anupam at 00:53:38) Yep. Thank you, my friend.

(Joel Beasley at 00:53:40) Talk soon.