Anthropic Member of Technical Staff on Leading Forward-Deployed Engineers, Turning Down 2x Pay, and Why Leaders Are Becoming ICs Again | Amandeep Khurana | E310

0:00I don't want to be 70 years old, not very healthy, sitting on$50 million. That trade just didn't make sense to me. You recently joined Anthropic and you have a pretty amazing career. I know that you started as an entrepreneur and then you took the blue pill and went to Amazon and then you kind of took the red pill and went back to building. Amandeep Khurana, member of technical staff at Anthropic. There were a couple of opportunities that would have paid 2-3x. My heart felt pulled in this direction and this time I wanted to do it differently. It's true that now, like a lot of these talented professionals that are working in Frontier Labs are being treated like NBA players.

0:33And the cup packages that we are seeing are absolutely insane. And I just don't know for how long this talent war is going to last. To be free of a game, you have to first win the game. And only then you realize the futility of the game and then you're free of it. The thing that is missing there is... Hey, this is Carlos, CEO at Product School and your host on the Product Podcast. My guest today is Amandeep Khurana, who leads a team inside Anthropics' forward-deployed engineering function. He's also a friend, so this is a first. We've had 100 private conversations and zero public ones. His path is unusual.

1:05Engineer at AWS, solutions architect at Cloudera, founded a company that exited to Databricks, then five and a half years at Amazon, where he voluntarily gave up his management role in 2023 to go back to being an IC. The things we'll cover, what a forward-deployed engineer actually is versus renamed professional services. The three filters they use to decide which customers to take. Turning down offers worth two to three times more. And why? On winning the game before you can quit it and where that advice breaks. Everyone at Anthropic has the same title, including Carpathy. Let's get into it. Welcome to the product podcast, Amandeep.

1:40Thank you. Thank you. I'm excited to be here and excited to spend some time with you. You know what? This is a special and a weird one for me because you're my friend. And we've had a ton of conversations about life, AI, and everything in between. But we've never had a public conversation, right? So just want to acknowledge that. Yeah, no, thank you. Thank you. Thank you for the opportunity. You're right. It's going to be an interesting one. It'll be fun. And it is a different conversation for sure. So I wanted to, I mean, you've been a collaborator behind the scenes. uh and i appreciate that but i just wanted to bring you on stage especially because you recently joined anthropic and you have a pretty amazing career i know that you started as a entrepreneur and then you you took the blue pill and went to amazon and then you kind of took the red pill and went back to building but now with with anthropic so i want to understand a little bit more about that part of your journey yeah for sure you know maybe give some uh give some background context so i trained as an engineer, took up a typical engineering job back in 2010, 2011 timeframe.

2:44I went to actually work for AWS. At that time, what I found was I, while I enjoyed the engineering work, I enjoyed the hands-on engineering, it just wasn't, it wasn't enough for me. It wasn't really what I truly wanted to do. I got this opportunity to work at a startup to go work with customers and I had no idea what that entails. So I went and did. I took it on, moved to the Valley. The company was called Cloud Data. So I went, became a solution architect, did all kinds of customer facing and partner facing work there. Spent five years, did a bunch of different kinds of projects and worked with a bunch of different kinds of customers and absolutely loved it.

3:19Through that experience, I also learned just by osmosis, what does company building look like? And I had this kind of dream in the back of my mind that I want to be a business person at some point. And I don't know where it came from. I think it came from watching my dad build his business. And so I wanted to do that. I said, you know what? Now I'm in the valley. The game to be played here is to go build a business. So what does that look like? So I started kind of thinking about it. And then I found a set of problems that I got excited about. And I said, I'm going to go do this. And that's how I started my company.

3:54And this was 2016. Did that. And then that company eventually exited to Databricks. Awesome experience, lots of learnings, lots of ups and downs along the way. So that was great. In 2020, I found myself in, you know, COVID hit and I wanted to start another company, but just kind of didn't happen. And I said, I need to do something. So a friend of mine was at Amazon at the time. He calls me and says, hey, you're up to no good. Why don't you come here and help me build this line of business that I'm building here? So I said, fine, you know, come do it for some time. And once this is done, you can go do whatever else you want.

4:30I said, okay, cool. You know, whatever ideas I'm working on, I'm not quite working out. So maybe I'll just go do this and not burn through my savings. So I joined Amazon. And I ended up spending five and a half years there building different kinds of things from ground ups, ended up doing four zero to one initiatives, two of them in the AI space. And even there, the journey is actually kind of interesting, because went from being an IC to in a management role, owning both product and engineering for to then giving up on all the management responsibilities and becoming an IC product manager back in 2023 because I just wanted to stop you I want to stop you there because that is a pretty unique right like a lot of times people want to grow up in their career sure assume that that means managing more people and kind of did that as a general manager at Amazon yeah and then it seems like you chose to go back quote unquote quote, to be an IC.

5:21And now it's cool again, right? But I think it wasn't that obvious or cool in 2023. And I want to understand more about that rationale. You know, it wasn't, you're right. It wasn't cool. It wasn't something that a lot of people were doing. I found myself in a place that there was this opportunity in front of me, which was an IC opportunity. And it was in a domain that I wanted to learn and I did not know anything about. If there was a management opportunity that I might have picked that over the IC opportunity. So it was like there was that and said look I want to learn this thing I want to get involved and let me just jump in and just see what happens and let me learn at least because I'm not gonna learn a ton if I'm gonna continue spending a large chunk of my time in the org building org management work I want to be hands-on and so that was one second is I spoke with a few folks at AWS at the time who had done similar transitions who had gone from a VP role to an IC role and then eventually back into a leadership role.

6:18And I asked them about how they thought about this. And so they advised me with some perspectives. And so I took that on and I said, I'll give it a shot. I have nothing to lose here. You know, worst case scenario, I'm going to learn about this new domain or best case scenario is I'm going to learn about this new domain. And as this team grows, there might be management opportunities that I can then take on. So that's how I thought about it. Like AI wasn't where it is today. And for me, it was, I want to learn that thing. I want to be hands-on and we'll see what happens. So you tried. I think you liked it because when you left Amazon, you wanted to kind of either build your own business, build a new business again, or continue working as an IC.

6:59But then you didn't choose any of those two. That's true. Okay, so the story there. So I do this thing at Amazon and five and a half years in, And I said, okay, I'm ready for something new and I'm going to go build a company. So I made up my mind. I'm going to go build a company. And so I resigned from Amazon. And then about a couple of weeks into it, I get an offer from Anthropic to come in and be a part of the forward deployed engineering team. And so when that conversation was happening, I did say, hey, I want to, can you just make me an IC forward deployed engineer? Because this is just so much fun.

7:29I want to do this. And the hiring manager for that particular role, he refused. and they said, no, I want you to come in and lead the team. So I'll be a leader on the team. So I said, okay, fine. You know, whatever the company needs, we'll make it happen. So tell me more about your current team. What do you really do at Anthropic? Yeah, so Anthropic recently started a forward deployed engineering function. Now we had FTEs. We had people working with customers on implementations. So it's not, that concept is not new at Anthropic, but we formalized it as a function relatively recently. I am a part of that function.

8:05and I manage a pod in that function. So I've got a bunch of forward-to-project engineers that are working with a number of external and internal initiatives, just working with customers, working with internal projects, and really bringing AI into the world in a safe manner, right, so kind of aligned with the charter and the mission of the company. And these can be projects of all kinds. The internal projects could be, oh, there's a new product we're building, let's go try it out, to there's a new initiative that actually impacts the kind of the entropic operations. So let's go bring AI into that initiative to working with customers on identifying use cases that can transform the way they do certain workflows, certain processes, and then going and building those out.

8:51So kind of the applied side of AI. Some of the folks also get involved in a little bit of the research and the model side of things, but not in the training. It's still once the model is at different stages of its training, that's when the team will get involved. The term forward deployed engineer now is super hot. And a lot of the companies that are doing some sort of enterprise motion, they claim to have FDEs that go and help the client adopt whatever SaaS tool they are selling. You guys are like the pioneers of this concept, right? So I really want to understand what it really means to do forward deployed engineering right, as opposed to just renaming the traditional professional services.

9:32You know, I would not call us pioneers of this. The concept of forward deployed engineering has existed for a very, very long time. You know, like Microsoft is the OG of this. They used to do it several decades ago. And then the current incarnation of the term was popularized by Palantir back in 2006, 7, 8 timeframe, right? So 2008 onwards. So we're not the pioneers at all. In fact, we are learning from them and seeing how do we take all of that and adapt it to the business that we are building. The way I think about it is there is, you know, in any company, there is building a product and selling a product.

10:05Two things. That's it, right? Everything else is just a supporting function for these things. What does the selling of a product entail? There is the technical selling of it. So this is your typical solution architect, systems engineer persona that will come in and, you know, they'll do the demos and they'll show you what the use cases look like. and they'll walk you through the architectures and so on and so forth depending on the product you're selling. So it's technical pre-sales, they're tied to certain accounts or certain verticals and they're incentivized to essentially close the deal. Then there's the post-sales and post-sales could be one of two ways, right?

10:35It is I'm going to help you implement the thing that you bought or it will be now it is implemented and I'm going to be the service or customer support team. So something goes wrong you call us and we'll we'll help you fix it. So note that both these things are the product team builds something technical go-to-market both pre and post sales teams go and make the customer successful with the thing that the product team built. FDE is forward deployed engineering. It is actually so there's a couple of different things about FDE that are different from your traditional professional services which is the post sales implementation.

11:07A couple of different things. One typical professional services is I'm gonna I'm gonna give you my time and you're gonna give me my money right so the exchange is time and money and you tell me what you want me to do and I will do it for you. I'm simplifying it, but generally speaking, that's the way it works. One of the core parts of FTE is it's not that I'm going to just give you my time and you're going to give me money and I'll do whatever you want. It is that we are going to align on an outcome. We're going to actually, we might even spend time together to identify the right outcome for the business based on the overall goals.

11:38I'm going to align on that and then I'm in it with you. We're going to make it happen, right? So there's that. So there's a lot more ownership in that relationship as opposed to, you know, you just tell me what you want me to do and I'm going to be your implementation person for that. So that's one component. The second is, whatever work we do will feed back into the product. In Anthropik's case, part of it is in the research as well, because that's how you build the flywheel, right? So FTE as a function is very closely tied to the product and engineering function. And that's where the value gets created.

12:16Right. So every engagement should produce a couple of things for us. One, feedback into product. Two, it needs to create some sort of a blueprint or a playbook such that that type of an engagement, we don't have to do it again. Right. It's a first of a kind, tip of the spear problem that has not been solved before. we will come in and help solve that and then after that that becomes a template that either again gets productized or we give it to the customers to build themselves if that's appropriate or we may give it to partners to then go scale it up as appropriate right so there's a bunch of system integrative partners that will take those templates and then go say this is how you do this in a bank so that's how we think about it it is not we're not trying to build a professional services arm we're not trying to scale a professional services arm this is to essentially figure out how to bring this technology into the real world in the best possible way and then take all those learnings and templatize them and and kind of give it away either in the form of product or in terms of blueprints at the same time i imagine that this service is reserved for a certain level of client right because it's harder to scale such an elite team to be able to go into clients and employees type of solutions or like what what makes the cut That's a great question.

13:34So if we look at the definition, right? So one, is the work aligned with our mission as a company? So our mission of bringing AI safely to the world, but not just bringing AI to the world, also for the benefit of humanity generally. So that's one. Second is, is it a first of a kind? And will it open up a set of use cases or a vertical for us that currently is not available? and is there some signal that we think we will get for the product out of this engagement? So we look at those three things and then decide, should we lean in here or not? We actually don't look at, the FDE team does not look at commercials as the only lens.

14:15In fact, that is, these three things are much more important than the commercials. Where do these FDEs come from, right? Because if you go to LinkedIn, everybody calls themselves either AIPM or Forward Deployed Engineer or however they want to call themselves, right? But realistically, like what does it take to become a FDE at Anthropic? You know, I don't think I can say that there is one kind of persona that makes for a good FDE. We've got FDEs that were engineers. We have FDEs that were kind of researchers and they have their PhDs. We have FDEs that come from non-engineering backgrounds as well.

14:54So the background, while important, is not the only thing that matters. What really matters is mission alignment. Are you hands-on and building things? Can you be in front of customers in complex environments and actually guide them through the journey? Because we have to help them. And oftentimes, even we don't know the answers. So we have to go figure it out. Because again, first of a kind, never done before. So we're going to run into problems that we never anticipated. Do you have high agency? Are you going to go figure things out? and really try to disambiguate while maintaining the customer relationship, while maintaining the trust of the customer and having their back and doing it in a manner that is aligned with our culture and our mission.

15:36So that's really what we look for. I want to talk about comp packages too. I think it's a big elephant in the room. We're hearing how these frontier labs are throwing money at incredible talent. But at the same time, this talent is being poached from one lab to the other, right? As I hear you talk about mission alignment and like kind of thinking beyond just pure comp, like how do you think about retaining this type of talent? You know, we have not, as far as I know, we have not really had a problem from a retention perspective so far. Anthropica has been growing very, very fast and the stock rounds are worth more than, you know, what you got, definitely worth more than what you got three, six months ago.

16:19So, compensation has not been a reason why we have seen any attrition. Anthropic generally has not seen attrition from what I understand. Now, again, it's a very young company, right? So, we still have to keep doing what we're doing and keep raising the bar for ourselves. In the long term, this has yet to play out. But we are a very, very principled company. And so, people, when they join Anthropic, they're generally very, very aligned with those set of principles. and that is one of the primary drivers when people join us. And we also screened for that in the interview. Having said that, you're right that the comp packages are very generous in the labs and much more so than the FANG companies, if you will, which was not true five, six years ago, right?

17:08The FANG companies were the biggest paymasters, the healthiest paymasters. That definitely has changed now. I want to share a quick anecdote that you and I talked about. You gave me permission to at least put it here, which is, in your case, the other companies tried to hire you for even more money than what Anthropic offered you. And you are really one of a kind that chose mission over comp. I still mean there's a certain level of comp that makes you be comfortable enough, right? So it's true that now, like a lot of these talented professionals that are working in Frontier Labs are being treated like NBA players.

17:47And the comp packages that we are seeing are absolutely insane. And I just don't know for how long this talent war is going to last. Yes. So in my case, there were a couple of opportunities that would have paid more, even 2 to 3x. My decision criteria was actually very simple. My heart felt pulled in this direction. and historically I have not operated from my heart. I've operated from my mind and this time I wanted to do it differently and so my heart pulled in this direction and I followed it. That was the line of thinking. Now, why did the heart pull in this direction? It was really because I felt connected to and the mission resonated and everybody that I met and how they talked about how the mission and how the principles actually play out in practice on a day-to-day basis, that resonated for me.

18:36And so that's why I chose this direction. In terms of talent wars, right? So my line of thinking is, if AI is to play out, all the benefits of AI that we talk about, ranging from medicine to bio research to chemicals to everything, right? If AI is to play out, the talent, the understanding of AI, especially on the applied side, has to be taken to the market. and there needs to be diffusion of it. Otherwise, if it's only going to remain with a small set of people, AI will not play out at the scale of humanity. It will remain concentrated to a very small group of people or the impact that the small group of people can have.

19:16So I believe that it is our job and our responsibility to diffuse this information, to train people, to teach, to make it easier to deploy, to make it easier to build, and so on and so forth. And that's very much a part of at least Anthropik's line of thinking. If you start to think about open-weight models, that's another form of diffusion that's happening. These capabilities are not, some of them, are not just behind paywalls. You have to pay for infrastructure, but you have these open-weight models and you have choice available to you. So that's another form of diffusion that's happening. Same thing has to happen with talent.

19:51Now, if you look at historically, late 90s, website developers were, you know, you didn't have many of them, and they used to charge a bomb. Well, today you just go to any of these, you know, we're lovable in one single thing and it's going to give you a full app. So the diffusion happens in every industry over time. And it's going to happen here as well. I think I want to acknowledge as I hear you talk about the decisions that you've been able to make is that you earn the right and the privilege to make them, right? to not take an offer that is 2x what you are making or to decide to be an IC or then a manager and then an IC.

20:31Like a lot of people who are kind of fighting on their way up and they are the people, let's call it director level. They are like, yeah, right. But in reality, I really want that 2x package and I really want to be a CPO and whatnot. So I'm curious to know how you think Think about it from the perspective of someone who achieved certain things in life and is now in a position to be more free to choose what's your next move. I've heard this, Naval Ravikant says this, right? That to be free of a game, you have to first win the game. And only then you realize the futility of the game and then you're free of it.

21:07I think he's right, but only partly right. I think, well, he's 90 % right. The thing that is missing there is to realize the futility of the game, there are a number of other ways you can get there. where it's not just winning the game. You can realize the futility of your game in other ways. How I realized the futility of the game was, as I reflected on my journey, so I wasn't a director at Amazon. I wasn't a VP at Amazon, for instance. It's not that I had got to a VP and then said, eh, what's the big deal? I can just give this up. I was very much on that ladder of growth, right? So for that definition of growth.

21:38But I also found myself in this place that I was just not any happier as I grew, right? And it just kind of wasn't cutting it. And then I had a few experiences, which then made me kind of go through this thinking of, okay, I went from point A to point B, I didn't feel any different. I went from point B to point C, I didn't feel any different. What were the things that were missing for me at each of these points? And I started to see a consistency there for myself. It was that there was just not enough heart involved in each of those steps for me. And so that realization over several years, it's not an overnight thing, over several years, that realization then led me to this place of I got to at least give it a shot and see what happens.

22:19I owe it to myself to let my heart be the leader here and then see what happens. Now, it doesn't hurt that the company's entropic. It doesn't hurt that they pay well, right? So I'm not compromising on anything from a quality of life perspective. If anything, it's an awesome brand and a fantastic group of people. So I won't say that I'm free of the game. I think that is not the right characterization. I can see the futility of it, but I haven't freed myself up of it. Yeah. And the way I think about it is ultimately you have the freedom to create your own game. And sure, there are certain parameters in life that are predefined.

22:56And if you want to live in certain parts of the world, you need to make a certain amount of money. But at five at a company and maybe at some point you're like, what if I define levels, those levels for my own company or for my own life? Totally. Absolutely. I think what you're pointing to is incredibly important. We get, I mean, I certainly did. I got wrapped up in the game that was defined by somebody else. The rules were defined by somebody else. What good looks like is defined by somebody else and so on and so forth. And some of those things just did not align with how I saw the world and the life that I want to live, right?

23:33So for instance, I value health and community. If I need to be a VP at take any company, and I need to work 60 hours a week, I won't be able to pay attention to my community, my family and my health. That's just how it is. And you see this, right? I don't want to be 70 years old, sick, not very healthy, sitting on$50 million. That trade just didn't make sense to me. So that's an example of me kind of observing the game, evaluating it for myself and seeing, does this fit me? And it just didn't fit me. One of the mechanisms that I think Anthropic is using, which is very smart, is everybody's a member of technical stuff, right?

24:13So tell me more about it and how are people okay with, yes, using the same term regardless of level. I think it is, this is a culture thing and people self-select into it, right? So if I'm in a meeting with somebody who would have been a VP, let's say I'm at Amazon, I'm in a meeting with a VP. I am in a meeting with a VP, right? So their word carries more weight than my word. There is this implicit authority relationship there, right? At Anthropic, I'm in a meeting with, take Andre Karpathy, absolute celebrity, you know, so deep in the space and somebody, so many people look up to, including me.

24:52When he introduces himself, sure, I know who Andre is, but he's an MTS and I'm an MTS. So the titles almost don't matter. Let's assume that I did not know who Andre is. I was living under a rock and I never heard his name. I'm in a meeting with him and I'm having a discussion with him. Title says MTS, mine says MTS. I look up in the company directory, MTS, MTS. We're on equal footing in that conversation. So it's all about that conversation and what you're bringing to the conversation and adding value to the conversation and having a respect for the other individual, not because of their title, but because of their ideas, right?

25:29So, and so people just bring, people self-select into it. That's how it, I don't have a problem with the title. I'm an MTS. My people who report to me are MTSs. I report to an MTS. Yeah, okay, whatever. And this is a great way to grab this app, Amandif. Thank you so much for your time and giving us an insider view into how Anthropic is growing. Thank you. This is a blast and thank you for the opportunity. Thank you.