From Stanford to Silicon Valley: The AI Leader who scaled WhatsApp | Wael Salloum

Welcome to Breaking Down Barriers, where we hear from the tech leaders in the ecosystem. Now, I'll start with giving you an intro. It's a big intro. So, well, it's one of the most influential data and AI leaders to come out of this Silicon Valley in the last 20 years. He built and led WhatsApp's global data organization, scaling it from four to over 200 across multiple continents. While the platforms use a base doubled, and its monetization strategy was defined and executed. Before that, he led teams at Google and YouTube. He worked on core data and AI systems that shaped the products used by billions, including Google Suggest, Google Ads UI, and the live-room experience for YouTube on TV. He also became known as a leader that brought into rescue critical projects and rebuild high-performing engineering cultures. Most recently, he's advised in companies across the mean ecosystem and beyond, and investors, leading data and AI initiatives that reduce fraud, improved operational efficiency, and supported market expansion across the region, giving him a unique perspective on how mean a tech can scale across the global standards. Why this conversation matters? Well, if you're a founder, if you're an investor, see sweet operator, government, or anyone building product and technology teams, well, has seen what good looks like at planetary scale. He understands how to build and lead senior autonomous technical organizations, and he has a rare view on AI that's actually adopted inside real companies, not how it's marketed. So today, hopefully, won't be about the AI buzzwords, it's about getting how decisions get made, how teams get built, and how scale has achieved in the real world. So, welcome. I said it was a big intro. Well, thank you, Richie. It's always a pleasure, and I've always enjoyed talking with you. Yourself have done quite a lot. I mean, you created an entire ecosystem here that has, even though I've been so busy lately in traveling, but I've really enjoyed it when I have been able to attend, and it's been a beautiful, thank you. You have made the region a better place. We need more people like you. Thank you. I appreciate that. So I'd like to start with a human bit to begin with. So, obviously, you originally attend, well, people don't know, but you originally intended to become a math and algorithms professor. What changed? What was the moment that shifted you towards AI and data? Ah, boy, this is a fun story. So I grew up in, obviously, in the Middle East, my mom, I was born in Quentin, and we sort of survived the Gulf War, and my father is Lebanese, and Lebanon was a lot of wars throughout that time frame. In fact, I was born during the Israeli invasion of Lebanon. So my life was fairly, you know, my parents did an amazing job in keeping it extremely stable, but you were aware of the greater political context growing up in the Middle East and stuff. And I just happened to be particularly good at school, and I love teaching so much so that in high school, I would teach, I actually would run classes, and I would skip ditch school to go attend classes at the university and come back and teach my classmates. And I thought the best thing I could do to help the region was to teach just my natural skill set and what I really enjoyed, and I'm very attached to this region. I'm not political, and I really just care about elevating the region. So I went to Stanford, got lucky enough to give a full scholarship there, and I attended, and there weren't at the time very few people from the Middle East and undergrad. I was coming as a graduate student, and certainly none from Lebanon that I was aware of. And so it was in new exciting when I got to Stanford, I had to literally send them my textbooks, and they made me do SAT2s, ST1s, and TOEFL, and a whole bunch of exams to ensure that I could be there. And while I got to Stanford, I was completely blown away by the quality of education that they had. I had an eagle. I was naturally the best in classes in the Middle East, and so when I went to Stanford, I thought I could do that. I signed up for honors, computer science, honors English, honors math, honors physics. I dropped out of half of them, and half of them I dropped out because I was not qualified, but also because of that teachers. I had one teacher give us the entire wrong exam, and this is at Stanford, right? And so it just, the professors and teaching was incredibly important to me, and I became a TA. One of the people who changed the course of my life, we actually met on opposite sides of a protest. He was on one side, this was how I was on the other, and while we were sitting there and talking, we ended up discussing a math problem, and so in the Middle of this protest, what everyone else was waving flags, we got together and we started solving math problems. And then he sponsored me to become a PhD student, so I became his PhD student. His name was Gene Golub, one of the most amazing professors that I've ever had at Stanford. Others include Andrew Inc., include Bradley Efron, Tim Ruffgarden, but Mark Van Kohl, but Gene really changed my life, and I wanted to follow in his footsteps, and I'm like, I just want to do what he does, but I want to do it in the Middle East, so my goal was that. And then what happened was my uncle, my aunt died a breast cancer, and my uncle got sick with Alzheimer's. And their kids could not afford to go to college, and these kids are brilliant. So if they're listening to this, they man and lemme, they're brilliant kids, and they needed some financial assistance. And my dad called me up and said, hey, can you help them? And I was the only one who could literally leave my job and get a six-figure income wrap and wrap the bat, and so I did. I joined Google with an objective of going back after four years to finish. And that's how I got into tech, right? I was already in computer science and math anyway. There's more stories there on how that became other things, but that's how that works. Amazing. I mean, I love all of your stories and your journey, and I'm sure we'll get into some others. You saw firsthand how poor data leads to poor decisions in US Congress. So what is the decision you saw that changed your worldview as well? Yeah, so for the readers who aren't our watchers or listeners who aren't aware, in the time I was in college, I volunteered and I dropped out calls for a little bit to work in US Congress. I was Senator Hillary Clinton's legislative aide on foreign policy. And while I was there, I was part of, or I did not participate, of course, I was way to junior, but I was, they had declared war on Iraq, part two. And they used, I'll say, I won't cite them directly, but they used a study which George Bush held in his hand and argued that democracies will spread peace and that, therefore, we should spread democracy around the world. And I was, I read that paper, and it did not make sense to me. I had a lot of problems with it, but I had up to this point, been a math and computer science person with a minor in Arabic literature. I was not a data justice person. Nor was I a political science and so I could not, I did not have the tools to articulate why this analysis and research was not making sense to me, but it did not make sense. But it was used. And that, those things, I realized that, you know, oftentimes people will use data in any way that they can to explain whatever story that they want. And you see this time and time again happening. And so I actually left there and I went back to school and switched into studying the applications of data. I worked with Tim Rothkarting on Game Theory on, I saw if it's better understand how to understand data and articulate the use cases and leverage it to make better decisions. And back then, before the rise of big data, so the decisions were making were much smaller scale. It turned out to be the right career choice. But you didn't know the time, right? And if you think about that, so across Google, WhatsApp, the Meenau ecosystem, what is one thing that's maybe remain constant about what you're personally trying to solve? A lot of people like to believe that they themselves are of high moral and integrity character. I also like to believe that of myself. I would like to leave the world a better place than it was that I inherited or I got. I want to build products that people can use that actually enrich their lives. And I want to help bring people closer together. So we live in a happier lifestyle. As a younger version of me, I did not truly value the amount of liberation that economic security and stuff brings. I was much more focused on this is a cool thing to build and use. Can we bring people together? Can I democratize information access education? Those are the things I was more looking for. As I've gotten a little bit more older, I've realized that also helping generate capital or opportunity is really a big benefit to them. But ultimately, what has always kept me going is less about money and success and much more about, I leave a good impact. I inspire people and is the world a better place now than it was before. We're making better decisions with it. I love that. So if we were to get inside your journey in the Silicon Valley, Google, YouTube, WhatsApp. So you worked on Google Suggest ads, UI, Chromecast, live-in-room, search, and ads. Which project taught you the most about scale? Ironically, WhatsApp was one of the top most of what scale. In Google, I also worked on sort of Hulk and Maps and geolocation data. That was fairly massive. Google search, there's all kinds of funny stories about Google search as well that I think I can share. One of which was Al Gore visited Google Office one time and me and a guy called Ahmed, one of the first engineers at Google, they had built the search queries that shows up in Pings. Well, he had built the original version. I only came to help him later. We did not filter out porn queries. That led to a big funny situation where Al Gore was sitting in front of porn queries going because porn is one of the most top Google searches. So a lot in Google search and WhatsApp in particular, just a sheer quantity that has happened. When you're at that scale, you have to consider different things. For example, if you have a 1% error rate or half a percent error, in most cases, no one cares about. But when you're at 3 billion, 4 billion daily active users, or I'll remember the current numbers of Google searches, that 1% scales up really fast. And then also, now, your millions and hundreds of millions of encounters are poor. That's a very big deal. All kinds of infrastructure cases shifting towards big data calculation, what that is at mean at that scale. One of the earliest things that we learned was experimentation. When you have small experiments, or once this was first developed, sample size was very hard to come by. So that's why you had sample size of five or ten. And then later on, you go to thousands or hundreds of thousands. But when you get to Google scale, and literally the smallest thing is hundreds of thousands, and you get to millions or tens of millions, then everything is statistically significant. This doesn't matter what you do. If it's even statistically significant because it's going to move in that range, the question is, how do you tell if it's really significant? And then the interplay, right? You assume every search query is independent, but they're actually tied to the user. Then you could search by user, but then the same users from day to day. And a query might be a refinement of a previous query. So when you do your experiments, a lot of the statistical statistics assumptions are independence between them, but they're not independent. So one of the favorite parts I had about Google and the shout out to Diane Tang and Amrnejmi and Hal Verian and others, is they were able to really re-invision what that world would look like. The design, overlapping experiments are sure they develop, they design the map produced with Jeff Dean. They conceptualize how to calculate statistical statistics at that scale, which was unproven math before. And so remember going back to the story of why did I not end up in a way? I ended up in the world's best PhD program at Google. I got to work on new innovative products. I got to work with incredibly brilliant people. And I learned so much. I still went back to finish my PhD and unfortunately my professor died shortly thereafter. So I never finished it. I went back to Google, but I have no regressive when it was the best job, the best sort of learning environment I could have had. It was such a great incubation period. And you became the person who rescued failing projects. What does that mean? What does a failure in engineering culture look like from the inside as well? So those types of companies? So Google in 2005 and 2006 is very different than it is today. And they were trying really hard and hired brilliant people and they were pushing on core advances. And a lot of times those projects failed. And that was fine. And in fact, Google rewarded it and encouraged it. In fact, I worked on a project for over two years with armor. It failed. And it was fine. No one kept your excuse. So failing was perfectly okay. The problem and what Google defined or other such companies defined as failing was that an issue occurred or it was not working, but people did not notice it. It was having impact and no one knew how to fix it. And it was a core bet. So we should be in all the ways we could not afford to fail. I mean, Google plus failed. And that was again, okay, Google moved on. So question is how do we resolve this and how did I became known about that was early on, I would notice problems in small systems. I noticed a problem in the map system where they were not connecting time zones correctly and stuff. And early on, more senior people than me noticed bigger problems. And I would look into it because I got curious and I would teach myself how did they catch it, what were they proposing. And one situation, a senior person found a problem, escalated it and then moved on. And the problem was not fixed. But I noticed it wasn't fixed. And I escalated again and it was not fixed again. And so I went and followed up a third time and our VP at the time is Shreeta Almoswani, who is the CEO of Snowfield currently. And this area blew up and I got assigned by this VP to handle the project and oversee it. It was the first big maker break project. I was so junior. I was very, very young in Google. I was only in Google for two years at that time. And for those that don't know what level in was that in Google? I'll be a four. So I was at level four. You were managing this critical project. I would project managing, not human managing. I was too young. I was too junior to manage humans. But I got assigned to oversee it. I essentially relocated to LA. I won't name the project here. And really helped turn that project around, formed a great relationship with the people down there. And then it was a huge success. And I was able to work with very senior people. I mean, the person in charge of the project was a director. And here I am coming as a level four and sort of overseeing it and people on the project were senior, several levels above me. So I got promoted after this project. And then I kept being assigned to hey, we have a problem. Go figure this out. This isn't going to slow. And so I joked that I was like the SWAT team for a while, sent in to save this stuff. And I got to work on tons of projects as a result. Hulk and Newfee and YouTube and Google suggests and Chrome and Earth map, everything. And it was a lot of fun. So I got to wide exploration of things. And I was able to really, I was glad I was able to go in, find a desk, turn things around work, work on my ambus person skill set. And I have learned a lot of technical skills. I really enjoyed it. I mean, it's interesting because some people like you say make or break, they could have broken their career potentially to some extent within Google, right? Especially at that time, that period in terms of the bar and how high it was and so forth. During that time then, who were the best operators or leaders? I know you mentioned a few that you observed in that era that have obviously gone on to do great things, but and what they did differently and how they then inspired you. I'll call out a few. And then I'm worried that others who I don't call out might be bothered. But there's a few really, really great leaders that I will call out. The first actually is Larry, Larry Page. Now I didn't work closely with him. He's, you know, we're not, we're not best friends, right? But Larry early on knew what he was good at when he wasn't, he hired Eric Schmidt to really build the company. He trained under him. A lot of CEOs of founders won't have that maturity. Right. He turned Eric, turned Google into a business, marked the same thing with Cheryl Sandberg, but he remained CEO and she was CEO. And Larry's case, you know, he literally stepped down and and reported to Eric and learned a lot. That was incredibly inspiring. That humility was fantastic. Amur Tang, Van Tang and Amur Nezumi. In fact, I even, I named my second child after Amur. Those two were extraordinary. extremely important in my career. Diane is brilliant. Incredibly, as long as Forsyth and stuff and moves fast and pushes hard, holds a high bar with a great deal of compassion. I've only broken down crying in my off-work twice. Once was in front of her, and she was a incredibly tough, tough person. I didn't have a good English language at that time. There's a funny story. I went by puppy at Google, and she kept telling me I was barking up the wrong tree, and I thought she was referring to my name, but she told me I was going the wrong path. I didn't realize what she meant. I did not know the idioms. She got me a book of idioms. She would sit down next to me as a senior director or later on the VP and teach me how to say "No" to people and "Time Manage." Just a level of care and compassion she showed. Armor was brilliant. I learned so much from him. I wouldn't go. He was really good at operational. He knew how to inspire people. He was able to identify and block things. And the key thing I learned from all of those people at Google was they knew their strength, then you had to play to their strengths. I thought that a leader had to be good at everything, but they weren't. I won't say now who was, but one of those people lost their temper regularly and with Yale. Another one was fairly reserved. The third one was highly academic, but they played to their strengths. They hired people that compensated for their weaknesses. They identified their strengths. They used this a lot and they worked a lot on their strengths. And they had this really well developed framework. So they didn't have to keep making decisions all the time. The early part of the framework was, you spent a lot of time in your 20s or 30s growing and try to, when you're young, you compensate for weaknesses in 20s, 30s, you try things. By certain point, you leverage your strength, not your weaknesses. So you go all in on your strengths and you go, - Keep saying that to me. (laughing) - And then they also had created structures. So there was in chaos, but not so much structure that has come all these over the rhythm processes. And they understood that structure was in place to remove decision fatigue. So I'm not deciding what to do or where to go. But if people did not follow the process, they didn't care as long as the job was done. So for example, they put in a agenda, here's a agenda we go through it. They minimize meeting overheads. But if people come in and say, "Hey, I have something important in ignored agenda." That was fine as long as it was in fact important. They prioritize care a lot. And so those were excellent leaders. Then when I moved to WhatsApp, I was inspired by Will Calfcart a lot. I had Chris Daniels and Matthew D. Mon, others, all of excellence. But Will was able to always remain calm, to go for WhatsApp in a very difficult stage, was able to really lead with vision. He was actually my age. He was the first senior leader that I reported to that was literally my peer. And I saw a huge gap between me and him. So he was inspiring to me. And like I wanted to achieve what he could naturally do. He was a product person through and through. I was a tech person. But his ability to make decisions. And I was so inspired by like a lot of people who are diplomatic and at scale, they played safe. But Will, I don't know, maybe because he was raised by lawyers, but I mean, was like, "No, this is wrong. We're going to sue them." And then he went out and made very hard, critical decisions in such a fast pace that I learned that you do not need to have all the decision, all the data to make a decision. And you really need to, and he took accountability when there were mistakes, when there was privacy policy and other errors. It took accountability. And I got to learn at the hands of exceptional leaders. And I was their humility, their strengths, their ability to focus, was what led them to be highly effective operators. And would there be some key fit to work with all of those people, whether it's some key things to play to your strengths that you installed in yourself, that you then installed into the teams that you built? Yeah. Now, I'm going to be very candid. I don't think I had much strength. I don't know what they saw on me. I mean, to this date, I share my feedback that I got from them with my team. One of the hardest things I ever got. This was the lowest point in my career. And I was shared with my team to say, look, this is the feedback you're getting. But this is the quality feedback that you have to write. But I guess what they saw was talent and potential, mostly. But I definitely don't think I had good strengths. I was unstructured. One of my earliest problems was I signed up for too many stuff. So I was too chaotic. And I learned from them. Like, I paid attention to their strength, and I adapted it. I took from Will, his eloquence, and his calm demeanor. I took from armor, his attention to detail, and his ability to build and focus on really technical insight to the world. I took from Diane, her visionary approach, and her ability to elevate the team. And I took from a borrower, her level of care and compassion for team members. So the way I approached it was, if I found something in a leader I admired, I would take an outturn around and I will impose on it. From Google, Google and WhatsApp taught me a lot. They taught me OQRs and snippets and things. And while now I don't bring that heavy level of process to the team because I want them to move faster and more agile, that structure I still really admire. So I would every week write, this is what I have to do this week, and this is what I did next week, and it keeps me focused. I block out focus time during the day. I tell the team to do it. I will help review their meeting cadence once a month to ensure that they're free and clear. I'm extremely clear in what I want the team to accomplish. I prioritize very heavily structured, constructive feedback. And I provide that feedback on a weekly basis. The team knows what it would have to improve it. I will call them in and I will-- I set an incredibly high bar of quality to launch. Recent example, for example, I've been helping property finder. And there's an exciting launch that's coming out. Hopefully it will be great. When I first came in, it wasn't the bar I would like, and I literally doubled the threshold for them to launch. And they've made it. How did you do that? So again, I don't want to disclose too much about this because it's coming out tonight. I want to let property finder have a small moment in the sun. But it's a really cool launch. And they set the certain coverage and quality bar in terms of data mapping and things. And it was below what I thought would be good from a user point of view. And so I pushed my team who was responsible for that. And I said, well, what do you think we can actually achieve? And so they came up with a number that they thought they could achieve with a lot of hard work. And I came back with a different number that I thought was actually the minimum bar we need to launch. My minimum bar was higher than their number. And I said, OK, you have to get to this. And they weren't sure how. So I sat and I worked with them on it. We worked through the coverage and the mapping and everything else we had to do. And at first, they didn't believe, but they started seeing gains as they were working on it. And at first, I was heavily involved. And I worked with them on a technical side and I reviewed them every day. But eventually, I backed off. And the team surpassed the expectations. And I think the biggest difference-- and what I learned from US versus what I've seen here-- is that people here don't always know what great is. And therefore, they don't know what they can possibly achieve. They're like, OK, this is good. I just look around. This is good enough. This is great. But-- >> You just showed them. >> Yeah. And this is a philosophy I've had ever since I was a child. This is this one that I think is definitely me. In high school, I remember giving a speech to my classmates saying, I want to reach for the stars. Because even if I feel, I'm still in the air. And the whole point was, I was not willing to accept anything less than 100% on every exam. And even if I failed, I still got good scores. When we built a recycling program in Lebanon, I was not willing to accept anything less than us having a nationwide recycling program that re-invents Lebanon. Clearly, as you can see, we failed. We've not done that. When I was doing a science project on the mechanics of Argile, I went really in depth. And I spent my life trying to learn. You keep setting a really high bar for yourself. And recently, I'm facing health challenges. So I set a bar for myself of losing 40 kilos in three months. I failed by last 20. So yeah. >> Yeah. As I said earlier, great. From when I saw you. [BLANK_AUDIO] - Well, and it's good to always have these targets in place anyway, and as you mentioned many times, you don't seem scared to fail, but you'd like to set the bar by to do it. - I am a lifelong failure. I wanna fill in everything I do. - So I just wanna index back on WhatsApp as well. You build WhatsApp's data and analytics or from four to 200 plus globally. How do you actually scale a team like that without actually breaking culture at the same time? - Oh, we build culture. We didn't just break it, we build culture. WhatsApp was an amazing company, but they were very anti-data. And rightly so. They wanted, WhatsApp wanted to be privacy first, and they believed if you logged data, you were not private. And that was a very valid belief because then if a government came after you, you had to turn over data and you couldn't do it. So when I came in, it was fairly antagonistic, right? I was not, well, warmly welcomed by WhatsApp. Well, I mean by the engineering team, not by the leadership. Obviously I was hired, so the leadership team wanted me. And I had never managed before. My first time managing and leading a team. And the first thing I did was I went out and I hired people who were both good at their jobs and resilient and I hired junior people. I went for junior people because I knew they would be willing to take direction and they would not always then challenge everyone and they would give the way to the engineers. And then I went and I solved the engineering problems. And you tell me what your biggest problem was and we're gonna solve it for you. And the biggest problem at the time was integrity. So here I was being hired primarily to help grow and develop WhatsApp and also Facebook really wanted access to the data and it was a whole separate fight with Mark and others. But instead I focused on solving engineering problems and we built and we worked with integrity team and we figured out how with no data and privacy where how to protect against child porn, how to protect against terrorism and that was a good win. And then I developed a data policy this is how you have to record to be able to do X, Y, Z and I gave the entire control over to the engineering team. And the data organization at this point was eight people and they were furious. We already can't do anything and now you're giving us like you're saying I can't get permission to do anything unless I ask an engineer first and I guess. And the CTO went out and spoke on that behalf and said we're implementing this and the engineering team was gonna run it and so everyone on the engineering team was on board because they had control. And then I just had my data team teached engineers why this was important. So at first it was very slow and they had to one-on-one convince engineers why they had to log stuff. And we logged it in a privacy-aware way. We logged it without any IDs or associates users, we'd not log a lot of information that we could have logged. And we logged things like if your app, what happens when your app crashes we can debug it, we logged, bandwidth issues so we can help improve things and we solved those problems. And then when we had to handle earning reports we instead of giving what's up data to Facebook we took Facebook's data and then all the matching ourselves and then turned over the number that Facebook needs first earning numbers. And the engineering team and eventually became so supportive of us that they turned over the entire program back to our control which was the plan all along. And at that point they started saying we need data people. We need this, we need that, we need to be able to move faster and they saw that the engineers who worked with data people accelerated and the engineers who didn't did not. And so the engineering organizations started giving us extra headcounts. They will give up engineers to hire more data people. And you know, four years later we were a big organization. - Oh, and during that time we'll see what's up doubled its user base and monetize. Like what was probably one of the hardest strategic decisions you had to make to obviously support that as well. - Oh, there was also during, that was also pre and during COVID. We had all kinds of problems and hard strategic stuff. And again here, like Chris Daniels and Will Calfcarrot and Matthew Dima were the key leaders who led us through that that transition. The first strategic decision I had to make initially was what data to collect and how to analyze. And here I took a lot of insights from Alex Schultz who's the CMO of Facebook on what to build and what to look at to help grow, grow WhatsApp and protect us and wherever things and what features and use cases we needed to build. And that insights provided us with good strategic decisions and bets to make. And of course we had the rising competition of telegram and signal. And WhatsApp was relatively slow in development at the time. And so one of the hardest decisions I had to make is that despite the success of the data team and accelerating stuff, it was now slowing launches because to launch something you had to do and implement all the logging and things. So we had to redesign new logging policy, which I worked with Ami and Nitinon and we agreed to launch some things without having all the logging in place. And you build it later on as you go. We had a huge challenge of around WhatsApp payments in India and what's the strategic thing to do. We got it wrong for over a year. And eventually we figured it out and we managed to launch it. Monatization wasn't easy. There was massive arguments between WhatsApp and Facebook about how to monetize WhatsApp. I can't go into greater details in that. But that took a long time as well. - One of the big arguments. - It was one of the biggest arguments. I'll tell you one, sorry, a member of my team did analysis that ran counter to a published Facebook statement by Mark on what we should be doing. And he says, "This analysis show is a song in the work." Now normally when you have an analysis that counteracts not only your own CEO but literally the CEO of your parent company, you run it by your boss first, it's a courtesy and then you publish it and we get to this thing. But no, he goes ahead and publishes it anyway. And I get called into this office saying, where does it come from? Why didn't you tell us all this other stuff? And I am like, well, I help my member with this with the right call, apologies, any lack of notifications on me. This is the right analysis, which we rethink our strategy. And to Facebook's credit, they were very open to these kind of conversations. There weren't surprises, it wasn't an antagonistic thing. But still, to counteract a publicly directed strategy, usually you'll be like, "Hey, wait a second, "I think this is against that, you run up the chain." And then you give them a chance to say, we've now changed strategy on to X, Y, Z, right? You don't have this massive, right? And so I publicly took the hit and I turn around to him, I said, "What the hell are you doing?" (laughing) Oh, to it. It's like, but you know, it was very fun to do. And again, what's up in Facebook? We're open to this kind of environment discussion. So it was okay. And those were massive debates, massive strategy, as well as a lot of what happened with the NSO organization that we end up suing. Those were really hard strategy decisions. I remember a moment where we were in the government of India and the guy was trying to understand how we could send messages back and forth about knowing who the people were. And that's just how we built this end-to-end encrypted. We don't, you know, but that part was extremely difficult for them to understand. It was a wonderful scene in the ministry. I don't know, there's a great moment. I mean, on that, like, well, on the privacy aware data and there was a core philosophy around it, like what's like the misconception that people still have about privacy versus personalization? I actually wouldn't call it a misconception right now. It's because, well, because right now, what might maybe is, okay. So a lot of people believe that to do personalization, you have to record a lot of data and has to be personally identifiable. And that's just not true. Absolutely, if you record a lot of data at the personal scale, you will get a far more better personalization. That's not an argument. It's easier to do and it's more effective. But you can get incredibly important personalization without having to record everything about you and all your backgrounds and all your use cases. And you're also able to do it in such a way that we don't have to, we don't have to know about you. You can create a zero-trust architecture where me as a company does not actually know your details, but your details are stored somewhere and then my model can predict based off of your details. And that's the part that people are missing. It's just that they store all this data and they keep it in. And I'll say this region is not the best on-premises. privacy will come to that. - Okay. (laughs) - Last question on this before we move to sort of the current times. You've sat in the room where decisions affect billions of people. What's the emotional weight of that? - Oh, really huge. I've had to make decisions that affected billions of people. I mean, ultimately, I guess to be very candid, I was always the date and the eye guy, whether it was a VP role or chief role or whatever. So ultimately, the CEO holds the pen. But in a lot of those cases, the CEO differs to their expert, right? So you make those decisions. And some of them are easy. You're deciding on the product that's being used by billions of people. And you're excited. This is a good feature. People are gonna use it. I want to launch it. I want to launch it fast. I want to launch it yesterday. Those are good emotional weights. Shutting down the product that's useful in an area 'cause it's not financial sense to the company, but literally millions of people use this hard decision. And that was, that was tough. These decisions that we had to, and again, I can't go into specific details, but one point in one of the companies that I was working in, actually, both of them, we had to decide about potentially walking away from billions of people. And that was really, really hard and massive decisions 'cause you believe in your product. You believe it's good for the world and how to use it. And then there are also emotional things. WhatsApp is used in war zones. It's used in, and so then there's real life. Now they're not affecting millions or billions. You're affecting 100,000s or so, but at a much more intimate level, like as in their safety's on the line. - Yeah. - Yeah. - Those were the ones I found the hardest. - Yeah. And I thought of this here. Now I wasn't actually put on the script, but thinking of Facebook, Google, WhatsApp, incredible companies, incredible people. I'm sure there's loads of stories outside of work as well of incredible conversations that you've had that you've met people. Is there any conversations that you had over the last 20 years, probably 10, 15 years ago that now today, you'll see in some of those conversations maybe come to light into what people was building the future of the world, the economy. Any particular stories that. - Oh, the problem is I don't know what sort of I can share. I would probably say my conversation with the Joel Lanzil. We don't always see eye to eye on a lot of political issues. But for those who don't know, Joel Lanzil is a founder of Palantir, the founder of Adapar, of ADVC, of OpenGov, and then much of others. It's a good friend of mine from college. And he. We don't see eye to eye on things. And I definitely have disagreements with him on a few things, but. We both had differing views on where the world was gonna go. I think he was right over me in those conversations, which probably explains why he's a billionaire or not. (laughing) That said, any particular things that he said we're seeing today? - I see, I wish I had no one with question on what I've asked him first. And look, this would be very candid here. Palantir has a mixed reputation in this area. And this region, rightfully so I have a complicated view of that company. But he. Remember, I grew up basically in the world of openness and transparency, and we bring people together, democratizing information through Google, connecting people through WhatsApp and Meta, very much on the B2C side of things. And Joel came from the B2B environment, more defense-oriented, high security analysis, helping solve enforce borders and the like. - Yep. - And then we can see where the world is today. And, you know, where they sit, yeah. - Where is it? - Quick, so just once before we move into the today narrative, obviously. So you're quick, succinct answers to this. I don't know, I went and said sorry, I'm a river of both. - No, no, no, just in these ones, so we're all good. So what is fundamentally different about building products here in Mina versus Silicon Valley? - I think that why I already said initially, people here are. I don't think they always believe themselves enough to build great, they settle for good. - Yep. - In your view, is Mina in a leapfrog moment? - Yes. - What is the window in who will miss it? - I think the window's only a couple of years left. And I think at this point, several countries have already missed it. I think only UAE Qatar and Saudi Arabia are in the running in this region. US, China, I'm worried about most of Europe. - Where's the first billion dollar AI consumer adoption wave going to hit in the region? - So candidly, there has to be the US. So it's going to come from the US and then. And honestly, the only place it could hit is in the UAE. So you're asking where? I think would you mean what? - Yeah. - Okay. So everyone right now is trying to compete in the LLM infrastructure space. I don't think that's where the billion dollars will be made in the region. It's going to be made on the application space. What exactly application I don't know? Very candidly, probably the first claim to that billion dollar would be taking an existing billion dollar company, their property funder or Caribbean or others, and retrofitting it to become AI first. And then now it becomes an AI. And I know some of the companies are already trying, Taliban and others are trying to do that. And I think that's a good thing. The EN is investing very heavily in this among others. But the first sort of native AI application that will hit a billion, I don't know yet in what space it's going to be. - And you're here now, you're in the Middle East, you've been advising, you've been consulting, you've been advising governments. When you look at the last 24 months, especially with someone with your background in terms of where you've actually been invested in real AI and data, et cetera, for a long time, what is real versus hype? - Well, first of all, I mean, I'm gonna start by answering a different question, which is, I'm extremely happy and grateful to be here. I moved here during a tough time in my period, in my life, to take care of sick family members. And so I took a step away from my career. And from there, I ended up advising both governments and companies, and in some cases working hands on. And so I got that experience. But I do wanna start by saying that even though I put claims on like they don't know what grade is here all the time, they have raw talent, and it is a really good, good region. We have the resources, we have the talent, we have the appetite. And you're seeing number of people coming in. I really believe in this region. I think this is, the Middle East is gonna go far. And I'm so happy about that, 'cause I believe it. So I wanted to say that, that was important to me. So what is real versus hype? Right now a lot of people are changing, chasing hype. And a lot of the gender of AI initiatives are feeling, I remember going to conferences and hearing people say, AI has only been around for a few years. I'm like, what have I been doing for the last 20 years? And they don't quite realize the, they're just retrofitting. They're like, they're taking, AI will solve all my problems, so they'll try to do an agent. It's like a chat GPT prompt on top of it, or they scrape a bunch of data, and then they use it. And it does accelerate a lot of stuff, which has resulted in a lot of things coming that have high valuation, but don't have much weight behind it. Most people don't understand the impact of AI on changing business modes. Most people don't realize the importance of your own proprietary data, and don't realize how to truly value your proprietary data. This region is way too heavily invested in consultants. And consultants are great. I end up working as a consultant for the last year, I guess. They actually know a lot, they know a lot because they talked a lot of people, and so they gather a lot of knowledge, but they don't have as much execution practice. And what people really need to do is just get their hands dirty and try. It takes about two years to mature an AI product to really understand it. You have to watch out for the Jevons Paradox. What is that for? So the Jevons Paradox is something that normally comes up in urban city planning, where essentially you have traffic congestion, so you build a bunch of roads, and then more, more people use the roads, more people use the roads, and then you end up with more traffic congestion than you had before, because now people buy cars, and more access and stuff. And basically, it states that the more efficient something makes, the more you actually utilize it, and the less efficiency gains you get. And so a lot of people think that they can, AI will self-fund itself. I will launch AI, somehow I will save a bunch of costs, and it will be more efficient than I was before. They're not accounting for the fact that people will be resistant to work in all of them, they will make them lose their jobs. And that if they have AI, their business will grow, and now they'll actually have more costs, and not the team ready to staff it and back it up. And you need actual expertise. If you use contractors and consultants, then you are building your IP outside, but no one inside actually knows how to run it or do it. And effectively, I mean, what has time and time proven is the basic parables of building a good business. You build a good team. You go team first, you build the right thing, you focus on your IP, you focus on what differentiates you. You buy or contract out everything else. None of it matters. If it's not on your direct line, it doesn't matter. You have very few priorities, you focus in, you align leadership on the consequences of it. So you do all that, you avoid the pitfalls, which McKinsey and MIT will say 85% of AI products are onshore ROI. Well, yeah, if you do those basics, you get it right. - Is that why you'd say that, you know, I was gonna ask you like, what do CEOs still misunderstand about building a data-driven organization? And does that come into some of the things that you're saying exactly that? - Yeah, so I will, a lot of it actually. So I think there's a few things that they, I will say they almost understand. The first is that true data governance, understanding lineage, the quality data pipeline, is a distinct skill set from traditional engineering. A lot of engineers don't have to build a scale production systems, but they never cared about the pipeline of data internally. They have to learn it. In order to really build, you cannot have a singular AI team that does everything, it will work for some cases. But if you do that, you will end up with one team trying to inject intelligence to everything. And I do believe that we are more and more moving to AI. I think there's a lot of hype around a lot of things, but I don't think that the AI wave is a hype. It's, you know, you'll go through the plateau and it's overblown, but it's gonna be fundamentally changed the way we are, just like MoMiles did. - So what would you say is your framework then for evaluating whether an AI initiative is actually working? - Well, I actually do, I do three things. First of all, I define extremely key metrics and ROI around it. If I cannot align leadership around it, then it will always fail. Second, once we have the key metric in ROI and we know what the goal is that we're trying to do, I then try to build the team with the right skill sets. We don't get the right skill sets, it will fail. If it's a core team, if it's a core IP, I build it internally. If it's not core, like a chatbot or something, I'm happy to buy, right? I don't why spend your resources in time doing it. And then the third key component of the framework is execution, execution, execution. Really drive them hard, daily updates, hold a high bar, high quality bar, get your hands dirty, try things, look up online or different things. That's it. - And do you think like the companies that are wasting money on these AI transformations, all of these foundations that you've just mentioned, they're probably what they get. I'm bringing in the consultants that you mentioned lack that execution. - Well, before I get attacked by every consultant company and the ones I work with, bringing in consultants and not the waste of money. - Yeah. - But what they often get wrong is a lot of people are so desperate to chase the AI train that they will just slap any AI on top and they'll miss the underlying fundamentals. So in one company that I worked here longer than others or I helped longer than others, I focused a lot on revamping the fundamentals. I let go of 30% of the team, restructured them, really booked up their AI skill sets, gave space for it, slowed it down, the competitor launched an AI bot very quickly, CEO was very upset. When the AI bot was, sorry, recommending the company I was helping versus the competitor. So their own bot was recommending their competitor. They became very happy that we didn't rush in. And now they're accelerating dramatically an AI. And I do believe the fundamentals I put in place really helped them because they built on top of a strong basis. Same with everything, right? The foundations. So I'm just going to finish up on some leadership and culture before we move to the quick fire and finish. What are the wrong mental models that leaders carry into data and AI organizations? - Oh, so I think they still think of data support function and the AI is everything. They think of AI as a feature and not as an infrastructure. And they think of data as just give me data. They don't understand that now data has to live in production systems. It has to be easily accessible, has to have low latency in order for AI to thrive. You have to have a lot of it, you have to have proprietary data, you have to have good governance and security. The, I believe they also think of, oftentimes if they have a central data organization versus democratizing AI across the entire engineering organization. - Yeah. And what separates people who scale with the company, especially right now versus the ones who don't? - So we are in transformative period and I will say that even if you see someone who looks like their father ahead in the AI, there are few, there are not that many who have actually been in the AI for 20 plus years. - Everyone's in AI specialist. - Not only is everyone's in AI specialist. I would say that the people who differentiate themselves now are, go back to the skill sets I talked about early on, which was the thing, one strength I thought I had was I was very willing to learn and I paid attention and I applied myself to learn. I learn every day. I learn from those who are more senior than me. I learn from those who are more junior than me. And I think if you do that, especially now in transformative period, you'll be fine. - If you were to teach one lesson to every single engineer in leader, what is it? - Execution. There's so many people who I think, well, execution with quality. - Yeah. - What is a belief you hold about AI that almost nobody agrees with you on? - I don't know if no one agrees with me on it. I think Andrew Eng is aligned from his talks. I haven't talked to him in a while. He was my professor back in college. But it's not something I see often talk about here. So everyone here talks about how data is the new oil of AI and AI is going to be everywhere. But Andrew talks about AI as electricity. It's not, you know, electricity, you see, your angular electricity is everywhere. But it's much more a fundamental architecture now that can drive a lot. And now you have everyone from a small company to a big company can utilize it. I still believe supervised learning, traditionally AI is valuable over the next 10 years. Everyone now is all in generative AI. I actually think you should not give up supervised learning. But also you should treat AI as an infrastructure and have a cohesive AI strategy and not build this AI feature, to say AI feature, to say AI feature, is what I see here a lot. - Yeah. - So here we are, the quick fire round to finish up. So you can just give one word answers however you feel. - Most underrated technical book to understand data in AI. - Pass on this one. - Most overrated term people use when talking about AI. - Oh. - Generative AI. - One job that gets more valuable because of AI. - Actually customer support. - One job that disappears the fastest because of AI. - Customer support. - Both open source versus closed models. Who wins long term? - Close models. - The region that will surprise the world in AI innovation. - This region. - The Middle East now. GPU shortage, temporary or structural? - Improary. - One startup category you would invest in for the next five years. - Applications. - The application of AI. - You've led teams of 200 plus and beyond. What is the one higher mistake pretty much every engineer and leader makes? - Including me. - We were often too focused on hiring for our current needs and not willing to take our time to hire the right leaders. - Yeah. - I agree. One sentence, the future of work in a world of AI is, enjoyable. And last but not least, the WhatsApp boardrooms decisions that the public has never seen. Yeah, I can't, sorry. I can't. Try and squeeze that in. What can I get? Catch you off guard. No, no, I, I'll, I'll, I'll, I'll say this. Both Google and WhatsApp have mean a lot to me. They are seller amazing companies. I'll give a shout out to a couple of companies. Those two have taught me a lot. They care a lot about their users. We agonize in WhatsApp a lot, but everything, including if only 100 users were impacted or what do we do, but, but the skill and there's some things that we just cannot share. But this public look at the NSO situation, there were all kinds of things that came up around privacy policy. There was a situation that occurred in contentious regions. WhatsApp in Lebanon, and we, we agonize over how to react. And one of the hardest ones was having to decide not to do something and something I cared a lot about. And honestly, the best company that I've seen right now in Middle East is property finer. It's really, really good, good culture, good amazing, great place to work. Where is it going, property find? I like what? Oh, I, I'm extremely excited about their future. They have really re-envisioned what property will look like in a few years. There are a few launches that I know are coming out very shortly. In fact, there was one launch recently. It was an avatar that led to a thousand percent increase in needs. But yeah, they're they're treating AI seriously and they're revamping. And they have a really aggressive roadmap plan and I think it's a company to watch. And if you were to try to convince someone from the Silicon Valley or in Europe that is an AI player to come and join property finder, what would you say? Probably finder is the Google of old of the Middle East. Thank you for joining Breaking Down Various. It's been a pleasure. Thank you for having me. Sorry for not answering the book on the question of the book. Oh, we can come back to it. We can put it in the comments. Okay. Thank you. And I've got one more thing to add for our audience. So I've got some interesting for you and a quick ask right now. 72% of the people who listen to this podcast every week are not subscribed. That's totally cool. But here's the thing. Subscribing helps unlock bigger guests, deeper insights and the kind of episodes that generally make a difference to the ecosystem. So if you're taking value from the show right now, even once could you hit that button? It's free. It takes one second and it generally helps shape the future of this podcast. And who knows you might tap into the reason why we get the CTL Amazon next month.