A new profession that's exploding
Have you ever wondered why the AI systems that big companies spend millions or even hundreds of millions to buy so often end up unused?
The answer is simple: AI isn't something you buy, plug in, and use. Every company has different data formats, different business processes, and different compliance requirements. That's why you need someone who understands both the technology and the business—someone who goes to the customer's company and connects the AI to the customer's actual situation.
That person is a Forward Deployed Engineer (FDE).
In 2026, this role suddenly blew up. Google Cloud said it would hire hundreds of them, OpenAI set up a dedicated company to do this, and Anthropic followed with billions of dollars in investment. Industry leaders have even said this will be one of the most sought-after jobs in tech in the future.
What an FDE does, in one sentence
A Forward Deployed Engineer (FDE) is a "super technical consultant who can write code."
A typical software engineer sits in an office, follows the company's plan, and works on tasks the company assigns. But an FDE is different—they go directly to the customer's site and work side by side with the customer.
They have only one mission: make sure the AI technology the customer bought actually solves problems, instead of becoming a shelf ornament nobody touches.
For example, a law firm bought AI to review contracts, only to find it couldn't tell the difference between US and UK law. That's when an FDE goes to the firm, modifies code and tunes the system on-site, until it works properly.
The difference between an FDE and a regular Sales Engineer
Many companies have "Sales Engineers," whose job is to help sell the product. An FDE's work begins after the contract is signed—making sure the customer actually uses it, and uses it well.
| 销售工程师 | FDE | |
|---|---|---|
| Goal | Close the deal | Solve the customer's problems |
| When they show up | While the customer is still deciding whether to buy | After the customer has already paid |
| What they do | Demos, testing | Write code, do integrations, solve real problems |
| How success is measured | Contract value | The customer actually uses it, and it works |
| Time frame | Short-term, close this deal fast | Long-term, maintain the customer relationship |
So the people who sell and the people who actually do the work are two very different things.
What does it take to become an FDE?
When companies like Palantir, OpenAI, Anthropic, and Google Cloud hire FDEs, they look at five areas of ability:
First, strong technical skills. You need to be able to write code, build systems, and handle complex technical problems.
Second, people skills. You need to understand what customers are saying and grasp their real needs, not just bury your head in code.
Third, know your own product. You need to know what your company's product can and cannot do, and be able to explain it clearly to customers.
Fourth, ability to execute and deliver. Talking isn't enough—you have to actually build things and solve concrete problems.
Fifth, a long-term mindset. Not finishing one deal and running, but maintaining the customer relationship so they keep using it for the long haul.
Someone summarized what matters most for an FDE like this: if you can only write code but don't understand the customer, what you deliver isn't what they want at all; if you can only chat with customers but can't write code, you're a consultant, not an engineer. You need both.
What does an FDE's day look like?
Suppose a law firm bought an AI system to help analyze contracts. An FDE is sent over and stays for six weeks. Here's just an ordinary day:
8:30 a.m.: The customer's IT security department says the system keeps logs too long, which violates regulations. The FDE immediately rewrites the logging code, coordinates with the legal team to confirm compliance, and documents the solution for future reference when other customers hit the same issue.
10:00 a.m.: A bug appears—cross-border contracts keep failing clause matching. The FDE pulls up the backend data, pinpoints the problem in five minutes, and fixes the code before the customer's morning meeting. The customer's team never even noticed the system was down.
1:00 p.m.: Compiles the day's findings into a report and sends it to their own product team, telling them which bugs every customer will encounter and should be built into the product.
4:00 p.m.: A partner at the firm raises a new requirement: the system should distinguish between US and UK law. The FDE sketches the architecture on a whiteboard on the spot, confirms it's technically feasible, and promises a prototype within days.
That's an FDE's daily life—wherever there's a problem, that's where they go to solve it.
A new model in the AI industry
In the past, one company built an AI product, another company bought and used it, and that was that. Things are different now.
The new model has three layers:
At the top are the big companies like OpenAI and Anthropic, which build powerful AI models.
In the middle are AI startups, which take the big companies' models to customers across industries and build customized solutions for them.
At the bottom are the real customers—banks, hospitals, law firms, factories—they don't want the AI model itself, but something that solves real problems.
Those AI companies in the middle layer are essentially doing FDE work. They are users of frontier models on one side, and engineers for specific customers on the other.
Two kinds of people on one team
Well-run FDE teams usually split into two roles:
The first is the Delta engineer, defined by speed. They charge to the front line, rapidly prototyping and rapidly solving problems, not caring much whether the code is perfect—get the system running first.
The second is the Echo analyst, defined by depth. They know a particular industry inside and out, understand what a customer really means by "this isn't compliant," and what changes are needed to pass review.
The two must work together. Only fast people, without industry knowledge, will build something that's simply wrong; only deep people, without hands-on ability, can talk but not build—useless either way.
What tools do FDEs use?
FDEs can solve problems quickly because they have a set of professional tools:
For orchestrating AI agent workflows, LangGraph or CrewAI. For enterprise data retrieval and knowledge management, LlamaIndex. For big data and databases, Apache Spark and SQL. For model testing and monitoring, LangSmith. For deploying and running systems, Docker and Kubernetes.
In short, an FDE isn't a worker with only a hammer, but someone with a full toolkit.
A deep truth about business
For company bosses, the emergence of the FDE role has changed how they choose vendors.
In the past, bosses would ask: Which AI model scores highest on benchmarks? Which is cheapest?
Now they should ask: Can this vendor station engineers at my company long-term, to truly understand my business?
Palantir is the best example. It doesn't make consumer products, only enterprise data analytics. In Q1 2026, its revenue grew 85% year over year, with US commercial revenue up 133%. How did it do it? Not by winning on model benchmarks, but by sending people to customers and solving problems together with them. It's that simple.
Before you enter the field
This job isn't for everyone.
Former Microsoft executive Keith Ballinger put it well: an ordinary product engineer might take a quarter to feel the impact of their code, but an FDE knows within hours. You write the code and immediately see whether it can stand up to real-world testing.
Extremely fast-paced, extremely high expectations, extremely intense. Not for everyone.
On compensation, Palantir averages around $238,000 a year, while mid-to-senior roles at OpenAI and Anthropic range from $350,000 to $400,000. And by 2026, New York has overtaken San Francisco as the city with the highest concentration of FDEs, because industries like finance and healthcare are all on the East Coast.
Final words
The FDE role matters because it solves a fundamental problem: good technology doesn't equal good application.
No matter how powerful AI is, if no one connects it to real business, it's just a lab exhibit. The FDE is that connector—technical depth, independent thinking, fast hands-on execution, and turning real customer feedback into product improvements.
If you love writing code but don't want to sit in an office staring at a screen all day, and you're willing to go to the front lines to solve real problems—then FDE might be the career path for you.