Cambridge-Absolvent entwickelt Iron Man KI

Cambridge-Absolvent entwickelt Iron Man KI

Show Video

- All right, so guys, this is a very, very special episode today because first, it's the first time we're doing an episode live in present, not on the internet. Second, it's the first episode with someone I know not from the internet, but from the real world. And it's the first episode in English. So there will be subtitles for the German audience, I'm hoping to find some service or some person to actually transcribe it and translate it.

But let's see. This is Sara. I will have introduced you to this audience already as a very, very smart guy. Certainly one of the brightest people I've ever had the pleasure to do homework with. So very welcome to my show.

- And my honour, thank you so much for having me. - So we were gonna talk about so many things today. We're gonna talk about Cambridge. Obviously, we're gonna talk about good old times. We're gonna talk about chess and all the interesting things. We're gonna talk about you being a CTO of a very interesting startup.

But there's this one big name on your CV that everyone's gonna be really interested to hear about. So, Palantir. - Yeap. - I mean, a lot of people recently know them from the IPO and talking about stocks and stuff but what do they actually do? I know it's like all very secret and like big analysis for big company, big tech but also secret services, governments.

How would you describe what Palantir actually does? - Sure. So yeah, lots of secrecy involved although they are with the IPO, doing a lot more publicity around what they really do. The short summary is big companies, governments, they will have a common problem of so much data spread around everywhere, but how do you actually utilise it and get some value from it? So Palantir, they come in to help provide the software that can pull in that data from all these different sources and help users get some actual value out of it. So that could be a government side, you have things like police departments, the data on crime and activity just as much even commercial.

Big data around automotive company, or the designs that they have or the cars that they have out on the streets, the results of those, any failures. Yeah, any industry you can name, there's data. It's obviously the big topic at the moment in the last few years. And Palantir's first goal is to help you put them into one place, organise it well, be able to take advantage of it while still keeping up with permissions and privacy, make sure you're not misusing that data by any means.

- So, and you were a software engineer for multiple years, I think? - Just about over two years. - Just over two years. And I believe if I remember correctly, you were forward deployed software engineer? - That's right, yes. - So forward deployed software engineer, that's like, I think if I remember, it's a bit like a software engineer required to also have social skills, something rather as.

Can you explain a bit, what's the difference and why did you choose that? - Sure, so maybe on the edge, so it's just skills wise, but the main difference is that you're client facing. Palantir is a product company, it's a software company. They're providing the same software for their clients.

But at the same time, you've got people working on the core of the platform that everyone gets but you still want software engineers out with the clients. You can do small tweaks for them, make sure the client can actually take advantage of the platform in the way that they need to. - So do you actually kind of wear suits and go into the client office and that'll be mainly your role but then you also do some engineering, Is it like kind of 50-50 or? - Yeah, so no suits, definitely no suits. Even though often these places are a little bit more suited up, so you always stand out a bit if you go there and we're just all in T-shirts. - So, you work for like a very traditional company, say, do like a contract with them and you come in as the kind of the IT nerds.

- Exactly. - Some data and software no one really understands. - Exactly, you just embrace that image after a while but you come in.

And like you say, there can be a spectrum for different clients but it will be some mix of the client interaction speaking to them, understanding their needs. And then the classic software engineering actually going and coding it and making sure it works the way they want. And then going back to the client, testing it, if they're happy you see what else they need.

Client's always want more after they see it, so you kind of go through this iteration. - Whereas a normal software engineer would kind of do just that, right? Working on software. What percentage would you say, do you spend actually doing software engineering and talking to clients, planning meetings? - It's hard to say because I think one part of Palantir is that they let people choose that percentage. So you get people who are very, but want to do the social aspect and talking, and they might even spend 50, 60, 70% just doing the talking. And then there's people who just wanna see the client data, but they don't really wanna do the talking and they're doing more like 10, 20% with the client. And then there's other people on the team that are doing more of the client interaction.

So it really depends, there's really no one answer. I think part of Palantir's like way of working is that roles are very fluid and every role is a spectrum. You can never say, this is the percentage between one and another. If you're a smart person and you can do the job, they let you kind of do it the way you want to.

As long as the client's happy, you do it how you want to. - Which makes a lot of sense in this kind of the modern approach and yeah, that's kind of what you expect from a Silicon Valley tech company. - Exactly. - Although, it's not a very typical Silicon Valley tech company. We're gonna talk about that as well. But definitely that kind of culture seems like that makes a lot of sense and it's similar at my job at Facebook of course.

So, as a forward deployed software engineer, you go to the client and I remember you were travelling actually into other countries. - Yeah, so those few travels, got to go to places like Australia. - Not too bad. - Not bad at all.

The biggest one was Denmark. I was there for about a year and then a few smaller countries on and off. - So you spend a year in Denmark, basically living in Denmark and just working in Denmark for Palantir? - Yes, Palantir provided an apartment for the year, lovely apartments. You choose when you wanna come home, they pay for the flights to go home. So you go for the weekend, come back on Monday and yeah, you work there and then come back for a weekend at home. - Wow, I mean, its a bit like a consulting life, so maybe consultancy, although it seems like you do very long projects.

I feel that sometimes I'm not a consulting expert but sometimes they do shorter projects. And this one year in Denmark, sounds quite like a long term thing. - Yeah, I mean, that was a multi-year project. I only stayed there for a year but there's people in the team going on and off. So yeah, I think unlike the consulting, it really is long-term projects.

They obviously get year to year, five year long licences with these clients. So different to a consulting both, yeah, you've got that travel aspect of a consulting. I think the main difference is even though they're multi-years, not many people stay on the one foot client for multiple years. They will move around and go to different clients. But again, depends on your interests.

There'll be people who are actually Danish, who live in Denmark and they just stay on Danish clients because, if they want to. Other people want to travel around more, go to different places. - You don't speak Danish, or did you learn it? - I learnt a little bit, enough to get food in a restaurant. But honestly, even then they were such lovely people and they had English for signs, even for menus.

To be honest, they had English... Yeah, they were just nice people. So they always accommodated. - Amazing, and what type of clients did you work with? I'm sure there's only so much you can say but what type of clients were? Because they are the kind of big companies, private clients but then what people really care about when they hear Palantir is like secret services, governments, police.

Did you do any of that? - Mine was mostly on the goverment's phase. Most I can't speak about. - So you're essentially a spy? You were essentially a spy? - Sadly, it's not as glamorous as you might think from a movie and things.

Definitely got to see some insights into governments as a whole and how they work. Spy might be pushing it a little. But I think like again, the Danish one, there's this public record of kind of their work with the police.

So as far as I can talk about it, like they work with the police, police department, work on crime. Just to be involved in that kind of thing is amazing. And just to see kind of the impact of the software can have on a police department's efficiency, yeah, you don't get to see that often. - It does sound extremely interesting and exciting. Now of course it can also be a double-edged sword, I mean, I think especially in the German media and the German social media.

But I'm sure everywhere, when you talk about working for the government, working for the police. I mean, Palantir does a lot of work for NSA, CIA, FBI, all of these big secret services. So not everyone likes the police, not everyone likes the government, not everyone would be, as I'd say comfortable working for secret service.

How did you feel about that? Was it something that you really wanted to do or were you a bit kind of... Did you think like you were a bit wary of this being maybe, I don't know, you don't know in advance what exactly you're gonna get, right? Like what are you gonna work on, sure? It's like, always like, we are the good people, but you never know. - Sure, there's unknowns there and obviously if you take in the US, you've seen all of the things there around privacy misuse in these government organisations. Firstly is like, I was super excited to it. A big part is how often do you get to see that kind of insight part of it? So that was definitely a big part of my interest. I think the second one was that again, I'm not Palantir more, I'm not selling them, but genuinely a big part of what they're doing is helping you get privacy first, and making sure people don't access data that they're not supposed to access.

When they access it, they use accessing for a particular reason. You've got all the thing about who's done what, when, and the trail of like how you've got to do something and how you use the software. And so, again, particularly US, you see a lot of the controversy around like people misusing data, using it for immigration and other things or like other spying on people.

- I wasn't even just thinking of about data. Like, I was more about like what they do. I also, like generally would say that most of the things that the government does or wants to do are like good things that I can probably support. But I mean, still like Palantir technology probably has helped kill the terrorists. - Supposedly, sure.

- It's likely they have helped killed terrorists. It's likely that it, I don't know, there was a lot of controversy when, coming to this ICE in the US, Immigrations and Customs something. - Enforcement. - Enforcement, right.

So that they would like, on the one end, they are like laws and rules but not everyone likes the idea of the government expelling people from the country. Not everyone likes the idea of military or killing people. So is this something that you worried about or do you know if it would have been effected by your work? Any of that kind of? - Yeah, it does affect some people. and obviously there's plenty of people who will join Palantir and say, I don't wanna touch government, is not my tech cake and that's totally fine, you go on something else. I can't say why, just plus it wasn't that, I don't wanna say not bothered by it.

But like I said, there's a lot of good in what these organisations do. Second is like actually getting a CA for yourself. You can make it on this judgement of, is it bad or not or is it just media kind of miss relaying what they're doing or spending it on the CF media. I don't know a lot of what happens inside.

So they have to like spend there, and obviously it sells as well. - Tell me about Facebook's image in the media and in the public and what's happening inside the company. There's some overlap, but of course of two very different worlds. - Yeah, very different worlds. So seeing it for myself working was a big part of my interest.

And third I think it's important to differentiate the like software aspect from the like policies. It's like democracy exists in society for a reason to work on those things. And you shouldn't like, it's not up to a software provider to work on like deciding what organisations should and shouldn't do. So I kind of isolate those two things out.

And I say, if you disagree with immigration policies and other things, those are things that should be lobbied against and worked on in politics. But that's very separate to the software. From the software perspective, it's you're focusing on, like, can we provide these organisations with the tools to be able to do things the right way? And whether they do things the right way is then the question for the politics aspect of things. - Yes, I guess still like choosing to work for them. You're kind of saying, like, I think that what the US government kind of does in terms of global policies is net positive.

Like there are some people that will be like, no, I'm so anti out of that, then I wouldn't do it, which I wouldn't personally. Like I'm with you there. It's important to hunt down terrorists, for example. Even though it can be uncomfortable to think that I'm working on like human beings being on their down, right? In general. I just wanted to ask you a bit to get your perception because this is always something that comes up.

And that honestly, is like, if you google Palantir, like literally things like one of the first questions I say, is Palantir evil? I think a lot of this because of the whole immigration area but also to some like, of course privacy and civilians and all of these. - Yeah, I think I'm not positive to democracy. If I believe in democracy, then I believe in like I should provide the right software.

- I like that approach. So you're saying you believe that the political system is stable enough to do good overall? - Yes, exactly. - So we should provide it with the best possible tools that we can and that's exactly what Palantir... - If that changes one day then should I change my opinion on working on these things. But as long as I believe in that system, then I have to believe that like you should provide the tools. - Right.

And of course that also kind of limits the governments you may want or not want to work with. And I know that Palantir is quite a patriotic company, at least from leadership perspective. Whenever I hear like Alex Karp speak, he does stress that Palantir is very US patriotic.

And kind of, American values and American politics is what they want to support in spreading the word, which of course to a large extent is freedom of speech and democracy and all of these amazing values. But then again there, of course, will be sceptics say, like do we really want to help the US try to become, rather remain such a global controlling power? - Sure. I mean, part of it is obviously I'm based in the UK. UK, US like, we all see they're allies and they work together. But at the same time, like no one non US citizen could ever see any of the US government work that's done.

So you're still not gonna get any of that overlap. And similarly... - Oh, so you have to be US citizen to be kind of involved in the projects for the FBI? - Yeah, you would have to be, or any US government work, they will have their own kind of policies around any thing they do there. So you're still like, just because you work at Palantir doesn't mean like you're working on those projects or even that you can work on some of them. So there is that divide there. Obviously there might be something you work on that ends up indirectly gonna one of those clients.

But again, any governments have policies and things and checks that you go through before you can work on that. And again, depends on the sensitivity of data and things where there who's allowed to see what data. - So now, since you excitingly work mostly for government types of projects, was there anything without naming a specific place or project or government, was anything that surprised you, anything that stood out to you? Like obvious is a very unique experience working for these types of organisations. Was anything that was different than you would've imagined as a citizen, maybe? - I have. I have to think about it. - Were they more or less competent that you would have thought? - I think more competence.

Actually, maybe I've seen both. But they have more potential and more competence than you might guess, I guess on the flip side, than they're always utilising. And that's where Palantir come in is and help them utilise what they could do. Yeah, it's a tough question. - These kind of like, questions are super hard to answer on the spot. But also like how involved were you kind of in closing the case and like end to end on the project? Like say you deliver your software, you help them use it, then do you actually get to experience like what they do with it? Are you actually like, oh, we caught this criminal, thanks to your work.

And you still part of that or is that kind of... - I guess it depends on the client and depends how they're working. Main things that depends are kind of, how closely you're working with that project and providing a very specific feature for them to do whatever they're doing and part, there's also sensitivity of data. So sometimes we're providing the software but we never get to see the data that goes on it.

And sometimes, like you said, you can dive into the data yourself and play with it. So yeah, it depends how hands-on you're able to be. And two, whether you have to be or not. Often where possible, like a client will share like we were able to do this, thanks to your software and that helps us or Palantir people understand kind of how the software is being used, and help learn from that and help them even more.

But yeah, it's not like you're becoming a police person while you're on the job. But sometimes you get to see some kind of visibility of what they've done with it and how they've done it. - Were there any like extraordinary security protocols, background checks that you feel like there was like, this was a big thing? Like there's some additional barriers to entering, joining Palantir or the projects that you normally? I mean, you always have background checks even when joining Facebook for example, but was it more? - For Palantir, joining the company itself, like I said, you can do or don't have to work on government projects, you can work on commercial and other things. So joining the company is the same as any other company. Working in the projects depends on the project completely, like you said. - So for some projects you might get vetted just for working on their project, even though you're already in a plea? - Exactly, yeah.

But that's part of the client side rather than from Palantir side. - Interesting. Your name must be all over the files of so many governments.

- No comment. I mean, no comment on any of that. But again, it depends what you're working on, how much you see.

Like I said, sometimes you can help on a project but never see the data. So you don't have to go through that same vetting. And people do want to go through that vetting because they are able to see that data might do, so.

- Sorry and I'm sorry for putting you on the spot. I mean, I can ask you all these uncomfortable questions. But to continue that theme, however uncomfortable you are, but I should have asked you before.

But are you comfortable with talking about salary and earnings and these topics, because there are a lot of young people obviously watching this and also interested just in the career part. We talked a bit about the work environment. Can you talk a bit more about kind of, because I talk about it from Facebook perspective, a lot like benefits and compensation and all of that. So you joined straight after a bachelor's degree, in London.

Palantir forward deployed software engineer, what's the type of compensation that you would get? And did you negotiate it? Were there any benefits? - So firstly, I'll come to benefits last. For salaries, grad salary, I think at the time you could get anywhere between kind of 60 to 80k starting salary plus options. Again, that may have changed since the IPO.

I'm not totally sure how that works. But at that time you could choose your package as well and they gave you three options, right? So you could choose more options and a lower salary or higher salary and less options. So you had that aspect there. Like I said it was 60 to 80k when I joined. I think that might have gone up since I'm not sure since then, like whether it's gone back down or anything's changed on that fund.

- That's a progression. You were there for two years? - Yes, two years. - I'm assuming with an increase? - Yeah, so you had two things.

Firstly, everyone who stays 18 months used to get a 20% bonus, a 20% raise, sorry. So which is obviously significant. And then after that, there was, I think it was a kind of default. Every 12 months after that was 5% raise. - So it's not performance-based.

- So that's like as a minimum, then after that then obviously you've got the performance aspects or if you change role and other things, they might change it more. But that was kind of like a default base that you're almost guaranteed. Say, if you're not hitting the mark, you're probably gonna get fired or not stay. But if you've lasted 18 months, you obviously doing well and you kind of got that as a default. So yeah, so those were the main salaries.

Like I said, that is changing or may have changed since I've been there because of the IPO, I'm not sure if they still give options. I know there are other types of stock compensation, so. - I mean, you were at Palantir for two years, you got stock options.

The IPO, that sounds like a dream for you. - Like I didn't expect. When I joined, the price or the value of the options or the strike phase, the five years before I joined, compared to when I joined, it went five X.

So I kind of caught it at this peak and then it stayed there for a while. So I thought I was like, oh, it's not really worth anything. I kind of I just missed the like big raise and then obviously IPO happened.

And then somehow it just... - What kind of multiplayer are talking about? - So when they started, it was about seven a half dollars which was basically at my strike five roughly. And so, yeah, I mean, you see it now it's that something $24 at the moment. I know it hit the lofty heights of $39 while it was still in.

There was lock up on the options. So you couldn't trigger them all yet, but yeah, it's a significant multiplier for something that at least when I left, I thought that's not gonna be worth anything, so. - And are you holding it? - No, I've sold mine. I think it has long-term value as a separate bit of your options. After you leave, you have three years to trigger them or they expire. So I had to trigger mine, otherwise they would have expired soon.

So that's why I triggered them. I think Palantir have huge long-term value. And if you want to invest, and I think they're good one, if you're ready to click it and hold for a few years. - Do you do a lot of like investing in stocks and all of that yourself? - No, I'm not. Like I leave it to any funds for ISIS and things like that, but I haven't done any investing.

Crypto is a bit too up and down high there for me. Yeah. - Okay, cool. I'm just curious because you are, I mean, a smart tech person. So some of them decide to spend a lot of time analysing these markets and some of them don't.

I'm also not really into finance and crypto as much as any of my followers would have hoped. I get asked a lot about my opinions on stocks and crypocurrency but I don't have that much. But it does sound very, very lucrative. - It can be for sure.

It can be downside. - When your company IPOs often joined. - Yes, IPO definitely is a big one. - Yeah. Very, very nice. And speaking of IPO we also need to talk about your startup, we will.

Because you left Palantir to join a startup but did you get a chance to interact with Alex Karp? I'm just curious. - Not directly. I think the closest was he used to run Tai-Chi classes.

He's a big Tai-Chi. So to the kinda California office, he would do a class in one of the main buildings and just the opening floor and anyone can join. - So you did take Tai-Chai classes with Alex. - I think I did once with one of those classes just to see him once, see what he's like. I know I've seen them even in the London office. Sometimes you come and he has, obviously as a big glass which is often empty but when he comes, he uses it.

And then I remember randomly turning once and he's there in kind of in his sports gear and he's just doing Tai-Chi in the room. Unexpected sites to like just turn around and see that there. But yeah, he's a character. He's a very interesting person. He does a lot of interviews. And of course now also with the IPO, they're sort of interested in what he says.

And some of the complains that he's very restrict and what he can say and like without manipulating the market and all of these. But yeah, he's also very critical of Silicon Valley. And I said before, it doesn't really seem like very typical Silicon Valley company, for example, in that it starts like work for military and secret services. Whereas Silicon Valley and big tech in general are quite left leaning and they're not that close to governments. I feel in general and Palantir seems like kind of definitely, yeah, the outsider in this community of Facebook, Google, Amazon, Netflix. Is this something you talk about at work or you kind of like? - I think it depends maybe, partly depends on your political interests.

And again, also depends if you're working on like a politically sensitive kind of projects, then you're probably gonna talk about it more. Yeah, I think that kind of politics is probably limited to the people working directly on some of those US government projects, more likely. Like I said, at the end of the day, the software company is just building software tools. So a lot of the time you're not thinking about...

- Facebook gets extremely political. It's like we have this internal Facebook, right? Which is called workplace. So this internal workplace, which is like for those who don't, it's like a copy of Facebook essentially, and become kind of an enhanced version of Facebook within completely separate database, obviously.

And we use it for internal communication. So we don't write emails, we ping each other on chat, on messenger and we do groups and posts and likes and shares, all of that internally. And every time there's like something political happening in the world, there's like this huge debate internally. It's quite interesting to see how it's obviously very, very left leaning, but there are a lot of controversies, and there's a wide range of opinions within the company. - Right.

I mean, Facebook's obviously a much bigger company. I remember like when I was at Palantir, I think it was about 2,000 people. - Only? - Yeah, so despite the magnitude of how much they do, they have scalability very well nailed. I think they have a lot more, they can do with it. But 2,000 people running everything, again, it might be bigger now, I don't know the numbers. But so you have a smaller, a much smaller pool of people to handle difficult and political things.

So I think it applied more to the US side. So I probably didn't get to see it as much as. I can't say it's hit too much on the European side. - Now, one thing that we left out was kind of the whole benefits and maybe working hours and that type of stuff. I remember being quite jealous because we were doing the bachelor's degree and then I stayed for the masters for part three and then you left to join Palantir, like with your etiquette or whatever salary and very nice working conditions.

- Correct. - What was it like? - Yeah. Okay. Well, firstly, benefits, I think part of is just like breakfst, lunch, dinner in the office.

If you're travelling then they're paying for hotels, apartments, all the food and expenses you spend, flights as well. Like they're doing everything. Their policy really is and again, this may have changed since IPO.

But at the time at least the policy was very much like we just want you to do your best work. Like you shouldn't be thinking about am I hungry or like how much am I gonna spend on this flight? Do I have to take a really early morning flight just to save some money and you're gonna be 10 times more tired? They're like, just take what you want as long as you deliver. And so yeah, you kind of, you got what you need to get stuff done.

So yeah, food is sorted for you, everything you need. Again, I think it may have changed a little bit since the IPO, and honestly, when you're reporting to shareholders how much you're spending and everything, that changes things. But at least at that time, it was very free.

- But you said you have to deliver. So do you work a lot mixed hours? - Mixed, I think, there are parts of them hiring grads. A lot of grads out of university or quite young is that a lot of them don't necessarily have like a manager's kids, family.

So you do get a lot of people working late hours. It's not necessary 'cause it's forced upon them. But there was some, a little bit of a culture of kind of one people cared about what they were doing and genuinely wanted to like just get things done.

But that also spreads a little bit, right? If you've got like all the rest of your team are like working later, like you can't be that one person that doesn't like, you kind of feel it. Even if no one's saying it, you can feel it a little bit. Having said that, there are plenty of people, do their hours, get things done and don't have to work crazy hours. But it does take a little bit of kind of I wouldn't say like willpower bit of kind of, you know what you have to deliver. And if you're just like fight a little bit of the, like it's not even peer pressure, it's just the societal pressure of if everyone else is working late, do I have to as well? I think you have to fight it a little bit I found, but it's totally fine that people can work their normal eight, nine to five kind of thing and get things done.

But I would say it's not the norm and you get a lot of people working more like 10, 11 hours, 12 hours a day. - That's interesting. That's definitely a bit of a difference, I would say to like Facebook working on us. I mean, there are people working a lot but I would say the majority don't work crazy. I would say 35, 40 hours a week, I would say.

- Okay, yeah. Again, it depends on different teams as well. I think I was obviously forward deployed, so I was more on the client side.

From what I saw, it was more on the client-facing projects because client work naturally has like need this deadline by yesterday kind of attitudes. And so that one can force it a lot more. - Yeah, there's one of the things by the way that I really like about my job, that I don't have clients. Like there's no one waiting at foot, right? When we change stuff on Facebook, like when there's a new feature or in my case, like integrity work being done, there's no one really waiting for it except when it's a privacy regulations.

That's definitely something where there can be that line but generally the work is not really time sensitive. - Right, yeah. - Which is obviously different when you're like trying to catch someone with the police or like I don't know, mitigating some incident or I know am making things up. - Sure, sure, I can't comment on any such cases, but yeah, the client work definitely has a little bit of that deadline based work. Obviously it ends up building really impressive things in really short spaces of time.

And it's not like every client all the time is like this. You might have it, like you say, something about the clients, so their world has change that's causing it or you're in like, you're at the start of a contract and they're trynna get going, or like they're behind some schedule. That kind of thing can happen. It's certainly not all the time, all year round but it is part of what can kind of burn people out is that, they're spending these long hours and then at some point it takes its toll. - Interesting. But it does sound like, I mean, it does sound like a very interesting job obviously and well-paid and a good employer but definitely a bit more, I don't know, maybe a bit more stress on the people, but more pressure and maybe in some other tech companies.

But then again, some startups also have very, very very stressful or high pressure work environments. - Exactly. Like I said, I didn't find it so bad. I think I had on and off times and if I had a stressful couple of weeks, I'll then take the next weeks easier.

And part of the Palantir flexibility is there like some days I'd come in at 10, 11 and I just woke up 12 even and because I know I've been up the previous nights kind of very late finishing work. There's no clocking in, no one's like counting your hours. If you're delivering, you do it in your hours, whatever works. - And is there a strong corporate identity? I mean, you came in here and it's my apartment with the old Palantir hoodie, so you're still wearing their stash obviously. But is there a lot of like culture of we Palantir offsite events? I don't know, team building? - Yeah, definitely.

I think I mentioned I love the reason people are working a lot is because they genuinely care about what Palantir doing. They're interested in the clients. They've chosen to work on particular clients 'cause they find those industries interesting. And I think Palantir is full of extremely smart people and so I think that also builds a sense of camaraderie that you're kind of like, you know you're working with smart people, you know you're building really special things and that builds that kind of team and identity that kind of, we can do this together. I think that's taken the company quite far, so far.

- Cool. And yeah, I mean that all sounds extremely interesting. One more thing I have to ask about Palantir before we close that chapter, are you a "Lord of the Rings" fan? - Am not, I've never watched "Lord of the Rings" amazingly. - You've never seen it, never? - No, I know, it's ironic, but I never got round to it. - All right. That kind of, yeah.

That completely ruins what I wanted to discuss, but no, that's fine. That's so surprising, you're a computer scientist. - I know. - And you haven't seen "Lord of the Rings"? - I know. - Interesting. Because the name is pronounced differently in the films.

- Is it? - Yeah, so it's like when there's like Gandalf and so there are these seeing-stones, right? In the film. But they say Palantir. Everyone says Palantir when talking about the company.

- Yeah. I have heard that pronunciation. - Palantir in the films. - That's interesting. I've heard that pronunciation used but it's not the dominant one.

- Obviously it's a fictitious Language. And I don't think talking has specified how this one's pronounced, that's what, like my superficial research on the internet. But yeah, I just thought that'd be curious and I was counting on you an IT nerd to help me out but, no, all right. - I'm afraid I can't help there.

- That's fine. Now, you had this job, which was very interesting, but then you left it for probably something that was at least initially paying a lot less because you you went on to become the CTO of a tech startup which is called Monolith Artificial Intelligence, which is another exciting buzzword. So what do you guys do? - Sure, so what we do is helping engineering companies and engineering, I mean, things like automotive, aerospace, packaging and building physical products, we help those industries design better products faster.

So take automotive as an example, 'cause obviously everyone has a car. They can kind of imagine what that's like; building a car, designing a new car. You have people work on design of the car, you have people working on just the body of the car, the wheels, there's people testing the crash test. Will they survive? How much damage will it take? Looking at is this manufacturable? How much will it cost? What's the CO2 emissions? So many different departments doing different aspects. And so you can imagine they're all designing their parts and then you need to test it somehow.

And so then they build a prototype, they go test it and then they see, oh, it wasn't as good as they hope because there's just so many things that are interacting. So, okay, let me go, let's go back, try some different designs, build a prototype, test it. No, it's not as good as we hope still, back and forth. - Right. - A car is incredibly expensive to build, the prototype costs a lot. So this iterations take months, years, costing hundreds of thousands of pounds.

And the question is how can we help them do that faster, cheaper, and kind of get to the better design in less time and less money? And the key part here is they're generating so much data. These automotive companies have been doing work for a hundred years, or over many years, they've built so many different cars and they've generated data on that time. What designs worked, what didn't, what would the performance of these different designs? And so we're using AI to learn from those designs and those results to then give them predictions. So if they have a new design, before they spend time building the prototype and testing it, can we tell them in advance, wait, don't waste your money and time. There's gonna be a problem here and here, you should try this design instead 'cause that will fix those problems. - Sounds like a more specialised Palantir, right? Something like Palantir is a lot about like, kind of getting information from all these various data sources.

And now you are doing predictions about building things specifically. - Yeah, the AI being on top is the probably the biggest difference. And so using the AI, but particularly obviously AI has been big in the last few years and there's been so much research and work done on there. One area that we're particularly focused on is AI for engineering, because engineering, you have a few differences compared to other AI like computer vision and things. One difference is you often have less data, depends on the type of problem, but your test, maybe they've tested like 500 different cars. That's still not a lot, it's not millions of cars.

So it's often lower volume data. Secondly, you have 3D geometries. So a lot of AIs looked at like image recognition. You have 2D. 3D is a whole different ball game.

And so there's a lot less AI that's been done or research that's been done in those areas. So those are two particular areas where engineering differs. And so where a lot of our AI is focused on is not just how we can do AI in general, but in these particular types of problems. - Right, that makes sense. And how do you get to become the CTO of a random AI startup? Like how did that happen? - Sure, by accident, mostly.

I didn't join as the CTO, I joined as software developer, kind of the first software developer. They'd done kind of a couple of prototypes in math lab and just like a quick react app. But I joined as kind of the first software developer to make the platform kind of what they'd done a couple of other projects with customers and they had learned a lot from it.

And they were like, okay, now how do we make something generic that we can use for all clients going forward? So I joined as that and then over time kind of evolved. And then I was kind of helping more aspects than just the software. And then out of chance just ended up becoming CTO.

- Just like that. Like how? Like at some point there must be like a conversation like, do you wanna become CTO? Like, yeah, sure, why not? Or like, how did it happen? - Sure. Yeah, without too much politics, I haven't had sub points like, you're kind of doing a lot of the CTO kind of all and then yeah. Yeah, it kind of made sense for me to be the CTO. - And how did you find the start up? Because often CTO like these chief positions are founding members and in this case, you were just hired into the start up.

How did you choose that one? - Again, a lot of luck. So when I left I gave my notice to Palantir and left. I didn't have anything lined up at that point. So I knew I needed a change. Again, Palantir is a fantastic company. Even when I left, I told them, like maybe I'll come back one day.

Like, I really liked what they do by just needed to learn. I wanted to learn things like team management. I don't want to do a different type of technology and I did not want to leave the UK. So I didn't want to go to any event Palantir's other work because it was in the US or elsewhere.

So I knew I had to leave but I didn't know where to. I was also kind of okay with just taking a couple months out. So I gave them my notice.

I just started looking at that point. So the accidental story was, I was initially speaking to an education startup. I'm very interested in education, I like teaching. So speaking to them, and then they were an accelerator called Founders Factory. And so I had spoken to the CEO who was kind of starting up really early stage.

I'm talking ideas on pen and paper kind of thing. So I had spoken to her and then I went to meet her in person. And the day I went, found this factory doing a kind of fair for all their startups.

They run an incubator and an accelerator. So I went to that fair and then I had to wait a few minutes 'cause it was early. And so I just walk around the fair. There's all these startups, no interest in any of them.

Like, not that they're doing uninteresting things, there's just no one I'm looking for, but I have to stall some time. So just go loop around a couple of times. And I see this company saying we're making Jarvis from Iron Man.

So yeah, for those who haven't seen Iron Man, Jarvis is like the AI assistant to Tony Stark who's kind of like suggest like, oh, you should try this design, this is what you should do. I'll go run the simulations and figure out. And then in the morning he wakes up and the new kind of suitors they're waiting for him. And so they're saying we're making Jarvis and I just laugh.

Honestly, I just laugh like, oh, this ambitious startup. Things they making Jarvis kind of thing. But I've got a few minutes the stall. So just go speak to them, just kind of and I say this as a joke. And they noticed like, it was genuinely as a joke I just spoke to them. I spoke to them and say, okay, I think the CEO was busy speaking to someone else.

I spoke to the COO at the time. She just kind of introduced me to better what they're doing. So, I was like, okay, cool. Still was talking to education startup. And then later I was like, okay, fine, I'll speak to the CEO.

I'll go meet him once. Again, still like nah, probably not but I speak to him. I was actually very sceptical of AI in general.

I find a lot of their use cases, I find either like, don't really make sense to me. So I think the example I always use is like a bank trading where obviously a lot of the trading automated trading now is using AI techniques. The thing I find hard to comprehend is that it's a modelling human behaviour. Like the stock prices move based on human behaviour.

Obviously part of it's signals from things like use and reporting of finances of these companies. But a lot of it is human behaviour and the last six months in Reddit has proven that. So with that... - Although the same goes for YouTube algorithm, for example, right? - Yes, exactly.

- It's a very successful AI based on human behaviour. - Exactly. So they work, I don't deny that they don't work. It's just something doesn't quite click for me. Then it was it's like modelling something truly mathematic.

Like obviously machine learning in particular. - You prefer it's a bit more of a specified space. - Like machine learning in particular is like essentially creating an equation, like some really complicated nonlinear equation that predicts some output based on some inputs. That's what it's doing.

It's approximating something, And so when it's approximating human behaviour, for me, I don't know something doesn't quite click where it's like, the human isn't some formula that you can put to it. So this is why I just wasn't that interested in a lot of AI based kind of work. And then the thing that Richard, the CEO kind of got my attention with was he was based like we're modelling engineering. This is like real physics. It's just so the physics equations are so complicated that you can't write them all out and simulate them directly.

It takes too long to compute. They're just like very complex, partial differential equations. They don't have nice closed forms. You can't compute them directly. So that's why he was like machine learning here genuinely makes sense. We're just approximating those equations as closely as possible.

That's what got my attention and I was like, yeah, okay, now I'm interested, like now you're talking. So that was the big part that kind of got my attention. And then the second one was that there was no product yet but he had clients. Like the time we talked to the people like rose voice Airbus, L'Oreal, like big names.

And it's just not a job there with no product. And I was like, okay, if you like the big part of this early startup is actually getting people interested in what you're doing. And like how many startups just die before they even start, because no one actually cares about what you do? - So I like that you came for Iron Man and you stayed for the partial differential equation. - Exactly. - Okay, very nice story.

This was Monolith. It was the startup you're with now. How has a Javas, like when you were like, you help a car company build cars. Like I'm not sure I get the Iron Man analogy.

- So, I guess the main thing is you want to guide an engineer as far as possible and to getting to a perfect design, but you can never take the human out of the loop there. For example, if you have BMW made cars for years, if you try and get an AI to generate a perfect car, sure, it can generate something, but doesn't look like a BMW. It just looks like a random generic car.

Like BMW have a style. There's a reason you know a BMW. You look out on the street, you see a BMW, you know it just because there's a certain shape that they use. There's a style that they have.

- And that's another Palantir parallel because they stress a lot that what they do is not just kind of doing like AI or data analysis but it's more of like providing humans with torts, right? So it's kind of like this this fusion, this hybrid solution where there's human intelligence and artificial intelligence or data intelligence and together they are there. It can be extremely powerful. - Yeah, very similar to what we have here. Human AI together, how can we put the two? And so, and then the Jarvis link is again, from the movie you see, like Tony says like, oh we had a problem with the cooling system, figure out how to make it work at these altitudes.

And then Jarvis goes away based on the data its gotten, run some simulations that are kind of figuring out, okay, at this altitude, what design changes can I make? And there's like exponential number of different design changes that can be done. But it does all the work of trying to figure out which design change is gonna make sure there's not a cooling or cooling issue. And so the same thing applies to automotive. You know it needs to pass a crash test in these scenarios, you know it needs to hit this performers, have this maximum CO2 emissions, but you don't know how to get there, 'cause every change you make changes everything about everything.

So that's where again how can a AI tell you what's gonna happen? And it's a mix of forward direction which is like, if I make this tweak, what's gonna happen? And then the reverse direction is, I wanna hit this performance, what tweak do I need to make? And so it's like a constant back and forth that engineer's doing of tweaking, getting suggestions, tweaking, getting suggestions till they're happy. - And how's it going? How is it? You didn't even have a product for a long time but still you're getting a lot of interest from potential clients. Now where's the startup? I think it has grown a lot. - It has.

So when I joined, so that was October, 2018, so two and a half years ago. So at that time, I think we should spend a few months maybe up to about a year kind of talking to these clients, just getting information, understanding their problem really deeply. And so since I've joined is basically like the platforms gone through iterations of existing.

And yeah, I think when I joined, I was maybe the fourth or fifth employee. So very early. I think now it's 24-ish, and got some big quotes that will be coming up currently, closing CVSA at the moment. So fingers crossed, I guess. - CVSA meaning? And that's by the way a topic on its own.

But we do have time. Can you maybe explain a bit how that works because I'm sure many people are interested like, how do these startups get from an idea to then getting maybe a round of funding and then the LEP or the series, series A, series B, IPO eventually. Like how does that work? Where's the money coming from? - Sure. So yeah, like startups can go through very different kind of styles of how they get money.

A typical one if you're following the like VC Fund raise kind of route is that, you come up with an idea that you're kind of bootstrapping yourself and just trying to get interviewing potential clients, trying to gather most info. And then people often get what's called like a pre-seed fund. So they get funding of kind of a 100k, 200k that kind of range on say small but obviously that's also a lot of money at the same time. And that can be from angel investors. It can also be from things like accelerator programmes and incubator programmes. - Which of course, yeah.

I mean, it is small for a company, but then, if you put yourself in the shoes, like you have an idea then someone gives you 200 grand for that idea, it is a lot. - It's a lot of money. It's amazing how quickly that you can use that all up. But it's some money to kind of get going. And so often people use that to then kind of get going a bit, test out some ideas, you hire a couple people to help build out a prototype, go do interviews. But you can see that actually, if you hire four people, then that money goes very quickly.

You've got a few months to prove whether this is a real thing or not. And your companies will die because either the idea is not good enough or they run out of money before they get to prove that it's a good idea. But yeah, hopefully, you are able to prove that it's a good idea, you get some like potential interest from clients. You've proved actually there's a market for this, there's some value in what you're trying to build. And then you then go try and get funding to actually like make this a real thing. And so that's typically like a seed rounds.

That can be anywhere between one to 2 million pounds. No prescriptively low but even as low as half a million but kind of up to 2 million kind of range. And so again, that can be from angel investors, but usually at that point, you're starting to get to VC Funds.

Maybe not the multi-billion VC Funds but the slightly smaller ones will start investing, again, if they like your idea. So a lot of networking needed, a lot of pitching. You can go through hundreds of pictures and everyone's like, ah, it looks interesting.

Come again in a few months when you've come more traction. Even though you don't have a few months, like you you need that money. It can be a very draining experience. And I think just as well as a really good idea and proven market, you genuinely have to have good persuasion skills as well and presentation skills and a good vision of where you want to go. But then yeah, if you hit that seed funds, then, yeah you have bigger the way, so you can hire more people, start building it out.

Hopefully, like as you get clients, you can start getting revenue and that's funding you along the way as well. And so that's why I say, like, it depends on the startups. Like some startups might want to be self kind of, what's the word? Like self-funding. Like their revenue is the money that they're spending.

The balance is basically, trends move very quickly. And whilst you could make small organic growth and it takes 20 years to build it, where by that time trends have changed, someone's beaten you to it. And so that's why the...

- I mean, there are a lot of very big companies that remain not profitable for a very long time, right? I think that even like Tesla was not profitable but making all of these cars and surely financing a lot of that from revenue but also getting a lot of investments when it technically they couldn't afford it. - Exactly. And you're basically making the bet that says like, I could do this really slowly, but it's a now or never kind of thing.

Like I don't have time to spend 10 years trying to slowly build this up. So once I prove, and this is an opportunity, you're basically, yeah, give me an investment, I'll build this 10 times faster. The investors, obviously, if they're investing in you, they want you to build quickly. So they giving you money so you can accelerate that growth and be the first to market or the first one that conquers that market. It becomes a bit of a battle where you're constantly like every time you grow, you're getting more funding, hiring more people. Then you need to make revenue in order to cover those costs.

And then you wanna prove yourself well enough that you can then get a bigger round. So you typically go pre-seed, then seed rounds, then a CVSA. Then after that, it's just... - So those are just names for funding.

Is there a categorical difference between these pre-seed, seed and series A or is it just more money? - I mean, it's slightly arbitrary to typically like I said, pre-seed, you're looking at kind of 50 to 200k kind of range. And so that one is usually angels, small investors. And then seed round is the kind of half a million to 2 million.

And then smaller VC is and/or angels. And then series A, you're looking at bigger VCs. For B series, again, arbitrary. But series B, series C, onwards is when you're getting to the bigger VCs and more of the like bankers finances kind of direction. And so in each phase, you're kind of also trying to prove different things. So at the early stage, you've got nothing, right? You're just trying to prove that actually the idea has potential and the market exists for it, and you're the right person to deliver for it.

And as you go and you start having to prove this is a good idea, and you're starting to go good prototype, you're getting traction, then the next series A, you might be like, it's a mix of vision for the product, but there's more around the revenue. And by this point you've had funding. You better be proving that clients are paying for this.

So you have to start proving revenue. And then after series B, C onwards, you really proving, this is like the money. - And where exactly are you? Are you selling your product to clients? - Yes, we are. So we have several big clients. - For how long have you been selling it? Because I think like, for how long have you not had product that you sell? - So I joined October 18. So I started building the prototype.

We used the first version with a client in January, February, 2019. Obviously it was worlds apart from where we are now but it was a genuine prototype and a lot of aspects of that still exists in how it was built. So that was like a three month project. Again, I'd say 2019 was a big learning phase for us where we were doing more three months, four months projects and just taking in as much information as we could. 2020 onwards is where clients were giving us year long licences. And there was like a more long-term kind of revenue.

But yeah, since 2019 it's been a product that people use. - And what does it mean for you personally? Like, it sounds very fancy and its really nice to be a CTO of a growing startup. Like you have various people reporting to you. So kind of in terms of hierarchy or in a sense much higher than you were at Palantir, how does the compensation compare? Is it already, would you say it's like worth it? It's very interesting of course.

But is it worth it financially or are you still ramping up towards that? - It depends. Like my conversation is more probably on the stock side than in salary. So it depends where you are in your life, whether that's a balance you want to make. I'm living at home, I've minimal expenses. So I really don't care about my salary at the moment. And although, it is certainly a nice salary and don't get me wrong.

It's not as good as Palantir but I've got stocks that are plenty in making up for it. If, when at some point I cash those in, then yeah, it might be worth it. Hopefully it's worth it based on the current growth. But your kind of, instead of getting compensation upfront in like the regular salary, you're balancing that with more stocks that you can then hopefully cash out later.

- Cool, yeah. I mean, of course, like it's a bit more of a bet, but it seems like it's going in an amazing direction and you are being part of that and leading it as a CTO. What's a typical day like? Every day is different, but what's your week like? What are the actual roles of you as a CTO of that AI startup? Do you use the actual coding or AI or maths? - Still codes.

It's probably one I'm trying to do less of than I do. And I have been able to. I think the startup environment is you're always under resourced. There's always more things you wanna do than you can afford to hire people to do. So for me I think the main things I'm juggling typically, is like the CTO, which is maybe more the strategy side, the alignment between different teams. That kind of aspect, long-term thinking.

And obviously I'm getting involved in things like fundraising as well but then also leading the software team, also being the dev ops engineer sometimes. Thankfully we are hiring for a lead software developer and a dev ops engineer to kind of cover those things. But until you can afford them, it's kind of me coughing in between those gaps. And then, but yeah, so then part of my time is on those recruiting as well.

Sometimes it's just making sure different teams are happy. Make sure like customer success team and the product teams have what they need. They have a particularly tough world that they're in between both the technology side of the product and the maybe business side of marketing sales, they're caught in the middle. It's a tough kind of vertical to be in. And so trying to make sure that they're getting what they need, they have the time to do what they need to get done.

That's often part of my time as well. And then honestly, sometimes it's bug fixing because the client needs it urgently. And I still remember the code 'cause I wrote the code two years ago. So I can go in very quickly do the fix, deploy it. I'm trying not to do that and thankfully software team are very good at kind of doing those things. But if there is something urgent and where other people are busy as well and I have to jump in, like, yeah, I'll have to get my hands dirty, it's not a problem.

- Now Essen, and your background obviously which we also need to talk about it's computer science and mathematics. And how much is of the work you're doing or the entire company is doing is just building software around existing air technologies or generally building kind of new AI technology. How mathematical is it? How much research even, how much science is it, or is it mostly just building wrappers around existing libraries? - It's both. So I would say like loosely our two USP's, one, making a software platform that's easy for engineers to use, and to do kind of AI kind of work without being data scientists. That's a big part of what we do, is making a platform that's easy to use. And that's like saying very software focused.

And then the second part is the data science work. So I mentioned there's a couple of big limitations where engineering is around low volumes of data and to do with 3D geometries. So on our data science team, they spend a lot of time doing research into the best AI matters to handle volumes of data, 3D geometries, to do research on things like how do you make an AI learn, not just like any model that works but like a physically correct model? You don't wanna a model that predicts that like energy doesn't get served. - [Announcer] A few moments later. - And of course doing that one time of the day where I record a podcast that was an Amazon delivery, we will have cut that out.

What were we talking about? - I've forgotten. - You're gonna be cutting out some more stuff. So we were talking about AI and maths and how research heavy it is, what the smart people do. And you have a background in mathematics yourself and you are very bright Cambridge graduate.

What are the backgrounds of the people working on that? Are they even like maybe physicists or pure mathematicians or phDs. - Yeah. So data science team in particular is who are doing a lot of the more mathematical base work. I think at least a couple of PhDs. You get a mix of kind of everything; astrophysics, electrical engineering, neuroscience.

Generally, m

2021-06-25 12:02

Show Video

Other news