0:00 I estimate that there are up to 30,000 job openings right now for a very specific job called an FDE engineer, a forward deployed engineer. And that would explain why five different people have asked me in the last week what an FDE engineer actually is. Well, I'm going to explain everything in this video. And this chart from Google Trends shows you what's going on. The red line is the number of searches people have made for full stack engineer over the last year, a typical engineering job. By comparison, the blue line shows you people searching for forward deployed engineer, FDE. It started to shoot up in June and it's been high ever since. Base salaries are being advertised up to $400,000, total compensation at the Frontier Labs,
0:45 probably north of a million dollars. It's easy to see why people want to hear more about the FDE engineer. So, let me take you through what it is. I'm going to demystify this in five steps starting with the backstory, where this role came from, then explaining the difference between how the term is used in theory and actually in practice in the market, the five different hats you wear as an FDE engineer, the job market, what it really looks like, looking at some jobs, and then the career path to get there. And as you can tell, this is a bit of a different video to my usual ones. There's going to be no coding. It's going to be just a tutorial. And if you like this kind of thing and you want more tutorials, then please do like and subscribe to this. It's how I know that you want this, and I will do more of this kind of thing in the future. But first up, the origin story. Where did
1:32 this term even come from? The term FDE actually predates LLMs. It was a role that was invented at Palantir all the way back in 2007. The CTO of Palantir came up with this. And the idea was, "Look, we can have engineers working at Palantir and assign them to our customers and actually embed them with the customers. Have them work in the customer's site and be responsible for deploying Palantir at the customer. That was called a forward deployed engineer, an FDE. And a couple of years ago, OpenAI decided to piggyback off this role description and describe FDE engineers as being AI professionals working at OpenAI that could be assigned out to OpenAI's clients. And so that became an FDE AI
2:19 engineer and it's what we now think of as just an FDE engineer. It's someone that knows all about LLMs, all about AI engineering, and is working at a client. And indeed, OpenAI announced a company this year which is all about embedding FDEs into organizations working on complex problems and demanding environments. That's how they put it in their blog post. But actually, the rest of the industry has now piggybacked off OpenAI's definition and you're finding that FDE is being used more and more. And it's being used in a looser way. It doesn't always refer to when someone with AI skills is deployed at an external customer. It's also being used generally to refer to someone who's using AI engineering techniques, LLMs
3:05 and agents to solve commercial problems. That is also how the term FDE engineer is being used. So to really press this point home, the term FDE engineer can be used in two different contexts. It could be referring very specifically to this narrow job of working at a frontier lab like OpenAI and being placed out to an external client responsible for implementing AI solutions at the client. Or it can be used more broadly to refer to any applied AI engineer that's looking to use AI to deliver commercial solutions, whether with an external company or internally at the company. So just to get super specific with you and so you have a road map for your learning if you want to do it, I would say that there are five distinct hats that you
3:52 might need to wear as an FDE engineer. Although, as you'll see, you don't need to wear all of them. Let me take you through the five hats right now. And the first hat's the most obvious one, software engineer. An FD engineer is a coder by profession. You would build, integrate, test, and deploy software. Many people come into the FD engineering space from being software engineers before. So, being a coder, that's one of the hats. That's a mandatory hat. You can't take that one off. And the second hat is my sweet spot. It is AI engineering. This is about knowing how to choose the right LLM and use LLM APIs, using agentic AI, rag retrieval, picking tools, using MCP servers, and then concerns like memory, state, context engineering, and the new harness
4:38 engineering topic. All of that AI engineering, a very important hat to wear. And now, you might be forgiven for thinking, "Okay, well, isn't that it? Those two hats? That sounds pretty important. What else could there be?" Well, that third hat is probably the most important of them all. And the third hat is perhaps could be called an AI solutions architect. But, it's all of the things here. And this is what really makes an FDE engineer an FDE engineer. And this includes activities like requirements gathering and discovery and evaluations. How do you measure that your AI solution is achieving your business objective? Then, data curation is about understanding the data you've got and the data you need. And the proprietary data you could obtain in order to have a competitive advantage.
5:24 R&D is about having the data science experimenter's mindset. Deployment is, of course, being able to roll out to production in a way that is scalable and secure and resilient. Observability is about being able to monitor what's going on in production, particularly with agentic AI. This is so important. Adoption is making sure that it actually gets used after it's been deployed and the change management around that, and then finally, all the aspects of responsible AI, including safety and security. So, obviously, there's a lot going on with this third hat. It's a big one. I didn't even mention things like orchestration, human in the loop. There's so much more I could say, but this AI solutions architect hat, that is the key difference between a standard AI
6:10 engineer and an FDE engineer. And now, for sure, you're thinking, "Okay, what could possibly be more than these three hats? Surely, this is everything you could possibly want." Well, there are another couple of hats, but they're more the optional extras. Let's look at them. The fourth hat is having business domain knowledge, being a subject matter expert in a commercial field. If you know about healthcare, if you know about financial services, this gives you a head start. This means you can work directly with a client and be solving problems for them at a faster pace than others. Now, it's absolutely possible to learn this on the job, and for sure, there are FDE engineers that get placed in a company like healthcare and learn about the healthcare business while they're there. But, if you already have some commercial
6:57 knowledge, it gives you this big head start. So, the fifth hat is a nod back to the original Palantir definition, and it's wearing a solutions consultant hat. This is someone that is able to represent your company at a client's. You're able to do things like stakeholder management, you're able to communicate with executives, you're able to be the bridge between your sales team and the client, and do things like identify new opportunities for new work. And also, you're able to take feedback from an implementation back to your company to be able to contribute to the road map. And you can probably see where this is heading. I mentioned a second ago that OpenAI has a more narrow definition of an FDE engineer, someone that gets deployed to an OpenAI customer. And typically, that would
7:43 require all five of these hats. That's what they're looking for. They're looking for someone that can do all of these things. But nowadays, it's more common to describe an FDE engineer that is someone that wears these three hats. Often actually working with the same company, perhaps working for a client, but typically, you're a software engineer, you're also an AI engineer, and you're an AI solutions architect. You're these three hats in one. That is the most use of the term FDE engineer. >> And of course, this isn't intended to be exhaustive. There are other skills that are very helpful and important. An example would be having DevOps, CICD, MLOps, AIOps experience would be very beneficial. And it almost goes without saying these days that having knowledge
8:28 of AI-assisted coding and coding agents like Claude Code is a must-have. And now to show you the size of the job market, let's start by looking at generally AI engineering jobs. People that can wear hats one and two only. It's a pretty big job market out there. The bright white numbers that you see here are the results of my analysis from running searches across LinkedIn and Indeed and so on. I found that there are somewhere in the region of 10,000 job openings in the US and then 30 to 60,000 globally. And the base salaries are in the range of 120,000 to 200,000 in the US. But I have to say, I think personally that this job board-based analysis understates the total number of jobs out there, perhaps cuz there are other jobs on career sites. I think, based on my
9:15 knowledge of the industry and talking to hiring managers and people interviewing, that there's several times that. There's more like 50 to 80,000 US job openings in AI engineering and several hundred thousand jobs globally. Take that with a pinch of salt, but that's my take. Next, let's look at the wide definition of FDE engineer. Someone that can wear hats one, two, and three. There are of course fewer jobs out there, but still quite a healthy number. I estimate 10 to 20,000 in the US, and then 20 to 30,000 globally with base salaries in the 140 to 230k range in the US. And then for someone that can wear all five hats, the narrow definition, the OpenAI definition of an FDE engineer, there of course the number of jobs comes down again. I could
10:01 find about 500 to 900 distinct jobs in the US in searching, but I reckon there's two to five thousand and maybe four to eight thousand of these jobs globally. The main point is that the potential upside is so much higher. Base salaries advertised in the 200 to 400 range, the total compensation, particularly at a frontier lab, would often be two or three times this and could well be north of a million dollars. And just to make this super concrete for you, I've gone into Indeed, I've done a search for FDE jobs in New York. The first hit, a forward deployed engineer with a base salary of 200 to 400,000. Let's look at it. And this job involves wearing all five hats. This is a full FDE engineer kind of job. You'll
10:47 see in what you will own that they actually mention the word embedded. You're embedded with these teams and it mentions that you'll work on longer term customer projects. It also says that you'll take your learnings from these projects and feed them back into the product roadmap. So it's all the five characteristics that make an FDE engineer. And it's 200 to 400,000 base salary in New York. You should look at local jobs to get a sense of the local market. And here's another example, a forward deployed engineer in healthcare transformation. This job does need healthcare business knowledge, so it does need at least four of the five hats. I don't see much signs that needs the fifth hat, so I'd say this is just hats one to four. And the base salary is in the 215 to 384,000 range in New York. So that's another job. They'll see one
11:33 here that's more cybersecurity related. Also got a similar kind of banding, and you can look through and see lots more jobs in this kind of territory to get a sense of the landscape. And I'm looking here at OpenAI's careers page filtered on FTE roles. You'll see a bunch of them in healthcare and legal. Let's click into one of them. This is a healthcare role in New York, and the description will tie nicely to the five hats of course. And if we scroll down, we'll see that the base salary is in the 185 to 300 range, but of course, with OpenAI, it's all about the total compensation and potential upside in the form of equity. And typically, a senior person that performs well in a role like this at OpenAI might be looking at north of a
12:20 million dollars of total compensation. So, here's the thing. If I've piqued your interest and you're considering a career as an FTE engineer, then the career path can simply follow these five hats. Let me show you. And you'd start this journey at the software engineering stage, and importantly, you'd be confident using coding agents like Claude Code, getting the most out of them, being productive without turning out LLM slop, and being accountable for all of your PRs. You'd progress to being an AI engineer, to wearing that second hat, by learning about LLMs and agentic AI, and delivering projects. It can be just personal projects showing your expertise in this field. And now we get to the big leap, wearing that third hat, the AI solutions architect hat. What
13:06 does it take to get there? There's two things. First of all, there's just some expertise to build on AI in production. But secondly, there is hands-on experience in a consulting role where you are working with a client to deliver solutions for them. And the best way to get there is through experience. If you can already wear hats one and two, then you can get the kind of role in which you are put in front of customers and you get to build that third skill set. So, you can do that on the job. That's the best way to get your third hat. >> And hats four and five, of course, are much the same. For domain expertise, it comes from real-world experience. Often when you start out in this kind of FD engineering job, you don't need to start
13:52 with the business knowledge on healthcare or accounting or finance or whatever. You take a job and you learn it on the job. And once you've built it, you can carry that forwards. Now you're a domain expert in that field. And the same very much holds true for the fifth hat, the solutions consultant. The best way to build skills in stakeholder management, in executive comms, in bringing back information about a roadmap back to the team, it comes from real-world experience in the field. So, if this has you interested and you want to progress, you want to take each of these hats in turn, then I would suggest here are your next steps. And the first step, perhaps the hardest step, is to invest the time you need in building up your own skill set to take on hats two
14:38 and three. Taking courses on things like LLMs and agents and AI in production. And of course, if you're looking for a course, then I know someone that has some great courses, linked in the description below. But I'm not here to sell my courses, there are tons of great courses out there. The hardest part is just freeing up your own time to focus on it. Now, you might not be expecting my second piece of advice. I know I said the best way to learn domain knowledge is on the job, but actually also it can be worth taking some time to invest in learning a business area, maybe two, before you actually get to the job. So, you can demonstrate that you are curious, you're interested, and you've got what it takes to apply AI to a business area. So, gaining some business knowledge, maybe in healthcare, maybe in financial services, it can only help you
15:25 as an FDE engineer. And thirdly, while you're still waiting to bag that perfect interview, you should be building all the time. Build, build, build. It's the best way to demonstrate your skills. If you're wondering what to build, I always advise people to stop by thinking of a personal or professional pain point. Something that really annoys you. Something that frustrates you that you would love to fix and then try to fix it with a solution that uses LLMs and agents. And then when you do that, you should showcase it. You should share it wherever you like sharing things. And if you'd like to, then by all means you could post it on LinkedIn and tag me and make it relevant to this video in some way, so there's a reason for it. And then I can weigh in and add some comments to show support and that way it
16:11 amplifies your impact and your expertise and it draws attention to a bigger audience. And I'd love it if you take me up on that. I'll put a link to my LinkedIn in the description. And you never know, maybe that post will come to the attention of a future client of yours or maybe even a future employer. Well, thank you for staying till the very end. If you like this tutorial-style video, then please do like and subscribe. That's how I know you want more like this. I might make one on open code set up and maybe oh my pie, a few other things that I'd love to show you. And I'll also of course keep making coding videos, too. And I hope to see you very soon for the next one.
