Andrei Savu – Principal at Sweat Equity Ventures. Ex-Twitter, Cloudera and PMC member at Apache Software Foundation. Sweat Equity Ventures is a Value Accelerator, a new kind of investor that invests expertise and time in exchange for equity. SEV is represented by a team of highly experienced operators; entrepreneurs and technologists; founders with decades of building highly successful companies, launching products used by millions, and creating billions of dollars in commercial value.
#5 Andrei Savu – Products & Technology at Sweat Equity Ventures on creating complete and contextually relevant products – Get Your AI On!
Ciprian Borodescu: I’m here with Andrei Savu, Principal at Sweat Equity Ventures, ex-Twitter, Cloudera and PMC member at Apache Software Foundation. I’m super excited, and it’s an honor to have you on this podcast. Thank you so much for being here.
Andrei Savu: Thanks for having me.
Ciprian Borodescu: You have years of experience in the tech industry, and I would like to invite you to take our listeners on a journey. First, tell us how and where you started your career.
Andrei Savu: I started coding in college and I think that’s also when I had my first job. The reality is I secured a job because I was kind of getting bored with college, which in retrospect, was a good thing – I got to learn a lot hands-on in a work setup.
Ciprian Borodescu: Was that the same time you became a committer for Apache?
Andrei Savu: No, that was a couple of years later. So, the job I started was a website that was kind of meant to be half portal, half-price aggregator. I think we’ve done something quite interesting there. We were, for a while, one of the top 10 websites in Romania in the eCommerce category. And after that experience, I took an internship at Adobe in Romania, for my end of college project, and that’s when I met a bunch of people there and one of them suggested that “Hey, you should do this thing called Google Summer of Code.” And that was my first touchpoint with the Apache Software Foundation. And I spent the summer part of the Google Summer of Code project, working on ZooKeeper with one of the committers there – Patrick Hunt was my mentor. And that’s kind of like where I started to get interested in the community, I started to understand more of how everything runs. And following that project, and following the internship I did at Facebook, I reached out to Patrick to kind of ask about other projects that are interesting at the Apache Software Foundation.
Ciprian Borodescu: And that’s how you became a committer and PMC member for Apache Whirr.
Ciprian Borodescu: Yeah, I started to work on Whirr pretty early into the project’s lifetime. I started to engage with the community there and became good friends with Tom White – one of the founders of the project – early while still in the Apache Software Foundation incubator. And I worked on this for a couple of years, I think for two years or so, helping with multiple releases and working through the process of taking the project from the incubator to become what Apache Software Foundation calls a top-level project.
Ciprian Borodescu: Yeah, yeah. How was that experience? How was the politics at Apache Software Foundation? I don’t have your experience with it, but I understand there’s a lot of politics involved.
Andrei Savu: I wouldn’t necessarily call it politics. I’ve learned a lot. So, yeah, in a nutshell, I learned a lot. And I’ve learned a lot about aspects of software development that are not immediately obvious. I learned a lot about what it means to do things in a way that makes it easy to build this global community that largely coordinates over email. I’ve learned a lot about legal licensing, intellectual property, and all those other things. I think, in my mind, Apache Software Foundation is an institution, it’s a very special group of people that get together and they are the reason why a lot of amazing technologies out there exist, ranging from the original web browser, but you know, in a way, they are responsible for the entire Big Data stack that changed everything over the past 10 to 15 years.
Ciprian Borodescu: Yeah, very well said. So back in 2012, you founded Axemblr, a technology acquired by Cloudera early in 2013, to bootstrap the cloud engineering team. I think that is a really interesting story. Can you tell us the story of that on how it came to be?
Andrei Savu: Yeah, absolutely. So it’s kind of like, one thing led to another. So I started to work on Apache Whirr and I was always interested in starting a business, but finding the right opportunity is difficult. And what I’ve seen while working on Whirr, I’ve seen a very active community interested in solving this use case, solving this problem of, like, how do I get my Hadoop Cluster in the cloud up and running? How do I start to process the data? How do I run a data pipeline there? And so on, and so forth. And it made me realize that, hey, maybe there is a commercial opportunity here because Apache Whirr was nice, but it was community supported, so if you had a problem, it was largely like, solve it yourself, or maybe someone will help you. But all of those are, like, more opportunistic in nature. So, I started Axemblr with idea that there should be a commercial play around Apache Whirr, and it makes sense to try such a thing. And the goal of the company was to take ideas from Whirr itself, keep advancing the open-source product, but also put it together and offer it as a hostess service, make a SaaS play out of it. And that’s kind of like where I started with Axemblr. In 2012, we spent around one year and a half building some interesting technology and talking to potential customers. And during the same time, I also had many different conversations with Cloudera where I also had many touchpoints from the open-source communities. And, in the end, early in 2013, I went to the Apache Software Foundation conference in Portland and, at the same time, I spent a week there, where I talked about a new project I was getting off the ground for the Apache Software Foundation, and spent some time with Cloudera in San Francisco at the same time, and this is where we kind of realized that the best path forward will be for me to join Cloudera and take all the ideas I had that I wanted to put in Axemblr and Whirr to another level with their support. And it played out very nicely.
Ciprian Borodescu: Yeah, I think that was an amazing opportunity. I mean, meeting them at the Apache Software Foundation, and then immediately just in one week or maximum two weeks, you got to this amazing deal. And did you have any colleagues helping you with Axemblr, or were you a sole founder?
Andrei Savu: I had a co-founder, but we kind of parted ways when this all happened because he had a bunch of other interests.
Ciprian Borodescu: Cool. So, I want to change gears and talk about your experience working at Cloudera. So, you spent there about four years, right?
Andrei Savu: Yeah.
Ciprian Borodescu: What was it like? What were some of the challenges, let’s say in the first year, and then throughout the rest of your time with Cloudera?
Andrei Savu: Yeah. I think there was a lot of ignorance on my side. I really didn’t understand what it means to join a company with 100 to 200 people. But I got lucky, in the end. I joined in 2013 – towards the end of 2013 – and I started to build a product within Cloudera that was taking mainly ideas from Axemblr, but in a different form factor. At Axemblr, we were pushing for building a managed service; at Cloudera, we initially thought that maybe the best form factor is not a managed service, it’s a packaged software that customers can install and run with it. So we built that and it’s been a very interesting journey. Like, we went from, “Hey, here are some ideas. Let’s build it out.” We had a launch I think towards the end of 2014. We launched the product at Strata in New York and we kept working with Cloudera customers for a couple of years, multiple releases, and so on, kind of getting to the point of realizing that, hey, this should also become a hosted service. So, we put some effort into building the Cloudera platform as a service offering, and so on. So going back to your question of how it was in the first year, it was an amazing rush in a way. I mean, I was just moving from Romania to San Francisco. I knew I wanted to be in San Francisco, and not necessarily anywhere else in the Bay Area, and I’ve learned a lot. I learned a lot about building products in a larger company that’s well funded, I learned a lot about working with product managers, working with customers in this enterprise set up, and also learned a bit about what does it mean to start to manage a distributed team. We ended up being one of the few distributed teams within Cloudera for a couple of years, and that worked really well for the work we were doing.
Ciprian Borodescu: Yes, a skill that comes in handy today, during Coronavirus.
Andrei Savu: Yeah, it takes quite a bit of patience, I think, to properly build a distributed team.
Ciprian Borodescu: And then, after Cloudera, you moved to Twitter. How did that happen? And what were your responsibilities there?
Andrei Savu: Yeah, so with Cloudera, I was lucky to basically see a company go from a couple of hundreds of employees through an IPO, become a public company, and I got to observe all of that, which was really interesting. I’m really grateful for that experience. But, at that point, after four years there, I realized that this market of data infrastructure is highly competitive. In 2017 it became clear that if you want to do a data project, the open-source toolbox was no longer the only viable option out there. Google was making a play, Amazon was making a play, a bunch of other players, so it became a very different, a very competitive environment and that made me think that I kind of wanted to move up the stack. So I wanted to get out of doing pure data infrastructure work to actually spend my time working with data. And that’s kind of what led me to Twitter, and more specifically to MoPub, which was in a way funny, because while doing Axemblr in Romania, I worked with a company that was a mobile customer, so I kind of ended up many years later going full circle, and ended up working for MoPub, following one of my first managers from Cloudera there. So, I ended up joining the data team within the MoPub Twitter business at Twitter.
Andrei Savu: And I joined at a very interesting time when there was a lot of change going on. And, as you know, change creates opportunities. And one of the major changes I faced when I joined MoPub was the fact that the main third party infrastructure provider for analytics for MoPub – the company called Meta Markets – was just acquired by Snapchat and that created this whole problem of like, “Oh, we need to move now, we need to take the data that was previously going to Meta Markets and create some alternative system.” And we had to do that very quickly. And because we had to do it very quickly, that unlocked budgets, so we got this opportunity to build a very large system for interactive analytics, and all the data pipelines in less than a year. So we went from concept to technology evaluation, sometimes looking at Google Cloud as an option and in the end, we decided that the right way to build this that will meet the requirements is to run it within Twitter. So, we ended up building this monster, like a 1000 nodes cluster processing hundreds of terabytes of data per day and providing the analytic support that the internal team at MoPub needs and the customers need. And I’ve learned a lot through that experience of what it actually means working with data at scale and moving the data around, and I have a lot more empathy for people that do that now than I used to have while at Cloudera because you realize how much identity that data has, in a way, and how much of a distinct asset it is that’s hard to move, hard to manage – you have to be very thoughtful through that process and so on.
Ciprian Borodescu: Can you describe a little bit the team you were part of at Twitter, while building all this stack?
Andrei Savu: Yeah, we were a pretty small team. It’s one of the things that I really liked about MoPub is that a relatively small engineering team was responsible for building and operating one of the largest real-time RTB exchanges in the world, processing well over 1 million queries per second, and all the associated data with it. But it was also that we had a lot of support from Twitter. So, we could definitely do a lot more than you could otherwise because of all the Twitter platform support that was available.
Ciprian Borodescu: So how big was the team?
Andrei Savu: The data team, I think, at the peak, we were like, six people.
Ciprian Borodescu: Okay.
Andrei Savu: And I think the group overall was around 30 or so, split across backend data and frontend.
Ciprian Borodescu: Okay. So now, you are a Principal at Sweat Equity Ventures, which we’ve talked about in the past and I find it very interesting. Tell us a bit about the investment model you’re proposing.
Andrei Savu: So yeah, the switch to the Sweat Equity Ventures came a bit surprising for me because I was not necessarily seeking a different opportunity. They reached out while they were recruiting for the portfolio companies. And my first reaction was, like, “You know, all these opportunities are interesting, but tell me more about you.” Because I found the model to be very intriguing. And the reason why it’s really interesting is because… I mean, where I’m going with this is that Sweat Equity Ventures is, in many ways, pretty close to my dream job. It’s an opportunity where I get to do engineering work combined with product strategy, business, and participate in various investment activities. So, as a venture firm, Sweat Equity comes into the market with this idea that the limiting resource for building companies is really the access to people, and not necessarily capital. And that’s the core of the model. As a venture firm, we take money from investors, we use the entire amount of funds raised exclusively for operational expenses that allow us to build an amazing team, and then we go in the market seeking investment opportunities, and we provide the services of our team in exchange for equity. Ideally, what we strive to be is the best partner a startup can have through a funding round. So you can imagine like a pre-seed company raising a Series A or a Series A company driving towards a bigger Series B style milestone – we want to be there to help them through all the challenges that are involved, make it easier to hit the various milestones, and at the end of the day, raise the next round at a better evaluation. And, in a funny-enough way, I think the work we do is as relevant today through the pandemic as it was before when capital was arguably a little bit more accessible because people now need to be leaner, so they need help from specialized people even more than before.
Ciprian Borodescu: Just out of curiosity, how’s your job now versus before? How much of your time are you spending writing code versus doing product or business?
Andrei Savu: I think it’d be more writing code, which is really interesting.
Ciprian Borodescu: Okay, I did not expect that.
Andrei Savu: Yeah, I get to split my time between a handful of portfolio companies and it’s a really diverse setup because I get to work on very different things, potentially, through the day – from things like process automation, computer vision, backends, data infrastructure, maybe write something like a browser extension, and so on, and so forth. So, I get to work on a lot of different things while also collaborating with founders on product strategy, on business strategy, on the go-to market, and doing some internal work for Sweat Equity Ventures around evaluating incoming opportunities or seeking new opportunities for the firm to invest in.
Ciprian Borodescu: That’s super interesting. And I know that we’re going off-track or off-script a little bit, but I’m just personally curious: how do you split your day or your week among different startups and different projects?
Andrei Savu: I think it’s tricky. I think I found a way that works. So, one of the first things I did when I joined the firm, was to put together a document describing how do I think, as engineers, how we should split our time. And it seemed to resonate with other people on the team, so I can tell you more about that. Basically, the idea I came up with was that in order to do engineering work, or for that matter, any other kind of work that requires deep focus, you need to have uninterrupted time, and you need to have a couple of blocks of uninterrupted time that are four hours or more. So that’s kind of like the core idea around how to partition your time. And the second thing is, like, even with this kind of partitioning, you still have to avoid doing too much context switching, especially from one business to another where maybe there is a learning curve, and so on. So, the way I’m trying to partition my time to ensure I can be effective, is I try to stick to two slots of high-context work; this is work that requires me to produce some kind of output, like I’m coding something, or I’m writing a document, or architecture or something like that. And these are high-context because they require me to either think through something that’s more complicated, or learn something new, and so on, and so forth. And I try to have two of those per week and keep them partitioned pretty clearly so that I can get to focus. Beyond these two high-context time partitions, I tend to also have a handful of what I call low-context things to do. And these are much shorter. So, rather than requiring four hours of uninterrupted time, these require more like 30 minutes and could be scheduled in chunks of 30 minutes – and these are places where the contribution I make comes more from my experience, than necessarily from the output I’m creating within that workstream. And this tends to be either working with the company as an advisor on some problem or more internally focused, like doing some due diligence or talking with a potential new founder, or like an EIR and so on.
Ciprian Borodescu: Yeah, it makes perfect sense to me. You’re spot on. And I am guessing it’s pretty hard to really be that strict with your timeline within a week.
Andrei Savu: It’s proved to be challenging for sure because the thing is, like, when you miss one of these slots, the task you’re trying to do doesn’t move to the next day; it kind of moves to the next week. That’s why you have to be particular about what you take on and how you manage expectations. I think what I’ve learned most through my time at Sweat Equity Venture in the past year is how important setting and managing expectations is and how important it is to communicate in a way that brings people along, that creates a strong buy-in. I think that, in a way, is a critical part of my job beyond the actual engineering work.
Ciprian Borodescu: Interesting. And how big of an engineering team do you guys have at Sweat Equity?
Andrei Savu: We started at two and recently we grew up to four. So, we are quite a sizable group of engineers on staff, beyond the rest of the team that also covers go-to market, executive coaching, people Ops, recruiting, sales, and so on.
Ciprian Borodescu: Alright. What are some of the AI startups in your portfolio and what are the challenges when coming in as an investor with some of these companies?
Andrei Savu: We have a few companies in our portfolio that do AI in one form or another. We have a company called Verta that does an AI infrastructure play. They are building kind of like an end-to-end solution for machine learning – and that’s been working really nicely. I think they have a really nice, unique take on the problem that’s led them to build an interesting high-quality product. But what we see more often is less of a developer infrastructure style play, but we see a lot more applied machine learning. And I think that’s going to start to be true across the board. As an industry, we’ll probably, at some point, stop talking about ML or AI as something special and simply acknowledge that it’s part of how we build software. There are things where you write code, and there will be components of a software product where you train models out of data. And these are not different elements, they are just different points on a range of how you go about building software. And to me, it’s a very interesting evolution of our industry, on how we went from handcrafting models to building out training machinery that can come up with those simply based on example data.
Ciprian Borodescu: Yeah, I totally agree with that.
Andrei Savu: And there’s a huge potential, I think, for applied machine learning. I think we’re barely scratching the surface of what taking the research that’s available out there means. I think, once it starts to get more broadly applied, we’ll see a really large number of products that have a human touch, in a way; they will have some behavior that’s more natural, they will be able to react to data in a way that’s more natural, they will be able to maybe show some kind of intuition in how the interface works, and so on and so forth. So, I think we are still very, very early on this journey. We are seeing some incredible results, but we barely scratched the surface.
Ciprian Borodescu: Yeah, still early. That’s true. Okay, so what are your plans for the second half of this year? Because right now, everybody is in this great lockdown.
Andrei Savu: Keep with the lockdown and keep with the work. I don’t know. I mean, I think in a way, it’s both a time of great challenges, but also a time of great opportunities and we should keep that in mind as we navigate all of this. It’s pretty clear to me that we’re probably not going to go back to the world that we had before – and in many ways, that’s a good thing. We need to go back to some other kind of world where we are more prepared to deal with these kinds of things. And I think it’s a great time for all the people out there that want to build things. There’s a lot of stuff that needs to be built. This pandemic has proved that most societies out there are ill-prepared to react to this kind of challenge and that means there is an opportunity to build things that will make it better. So yeah, let’s build.
Ciprian Borodescu: Nicely put.
Andrei Savu: That’s what I intend to do for the second half of this year: build some interesting stuff.
Ciprian Borodescu: Okay. So, we are nearing the end of this episode, and for the final special section on the podcast, lightning questions and answers, which is a series of fun, short questions that you have to answer really, really fast. Are you ready?
Andrei Savu: Yeah.
Ciprian Borodescu: Okay, so here it goes. Game of Thrones or Lord of the Rings?
Andrei Savu: Game of Thrones.
Ciprian Borodescu: Star Wars or Star Trek?
Andrei Savu: Star Trek.
Ciprian Borodescu: Okay. Now, I want to see this one: Joe Exotic or Saul Goodman?
Andrei Savu: Actually, I don’t know who they are, so neither.
Ciprian Borodescu: Okay, let’s see if you know these guys. Elon Musk or Jeff Bezos?
Andrei Savu: I like Bezos, but I wouldn’t necessarily work at Amazon.
Ciprian Borodescu: Okay, that’s an interesting answer. What’s your favorite movie?
Andrei Savu: I really liked Herr.
Ciprian Borodescu: Okay. Cats or dogs?
Andrei Savu: Cats, but I’m allergic to them.
Ciprian Borodescu: Okay. Favorite startup?
Andrei Savu: I don’t think I have one. I think I like different companies for different reasons.
Ciprian Borodescu: Okay. The last book you read?
Andrei Savu: The last book I read was Complete Family Wealth and what I’m reading now is a book called The Knowledge that talks about restarting the world after a pandemic, which is really interesting.
Ciprian Borodescu: Okay, that’s really timely. Open-source licenses: MIT or Apache?
Andrei Savu: Apache, but I’m obviously biased.
Ciprian Borodescu: Of course, of course. Hence the question. Romania or the US? Talking about bias.
Andrei Savu: Both. I think each country has strengths. I see my heritage as a Romanian as very important. So both.
Ciprian Borodescu: Awesome. I like your answer. So the last question, the bonus question. I know you’re passionate about quotes because I see them almost every day on LinkedIn now. So my last question to you is, what’s your favorite quote?
Andrei Savu: Yeah, this made me think about another book I read that talked to me as one of the very interesting books I read. It’s a book by Alan Watts called The Wisdom of Insecurity, a message for an age of anxiety. And it’s a really interesting book, it’s a book that talks about the nature of language, what does it mean to be present, and a lot of the hardship and natural anxiety that comes with us existing as humans. So, it goes like this. “The power of memories and expectations is such that for most human beings, the past and the future are not as real, but more real than the present. The present cannot be lived happily unless the past has been cleared up and the future is a bright promise.”
Ciprian Borodescu: Okay. This is deep and philosophical. We have to think about that before reacting, right?
Andrei Savu: Yeah, yeah. I think the gist of the book and one really interesting idea that I keep coming back to is accepting that we just don’t know. There is a lot more out there that we don’t know, and trying to make that go away with simple explanations is not helping anyone.
Ciprian Borodescu: Yeah. Or in other words, the way I would explain it is, like, there are some things that are in our control, but most of the things are not in our control.
Andrei Savu: Yeah.
Ciprian Borodescu: Awesome. Andrei, it was a pleasure to have you, and thank you so much for sharing your wisdom with me and with us.
Andrei Savu: Thank you. Thanks so much for having me.
Ciprian Borodescu: How can people reach out to you for ideas and comments?
Andrei Savu: I’m pretty out there on the internet. So, on Twitter or on email – my email address is firstname.lastname@example.org. So, feel free.
Ciprian Borodescu: Awesome. Thank you so much, Andrei.
Andrei Savu: Thank you.