What is a forward deployed engineer?

FDE stands for forward deployed engineer: a software engineer a company embeds with a customer to make its product work in that customer's real systems. The FDE writes production code there and stays until it works.

On the FDEBench board alone, 1,453 FDE roles from 799 companies have been posted since May 2026. Below: what the job is, a typical week, where the title came from, who hires, the pay and the interview.

What does a forward deployed engineer do?

A forward deployed engineer turns a product that works in a demo into a system that works in one customer's business. That means real data, real security rules and software that was written years before anyone said "LLM".

Compare it with a product engineer at the same company. The Illinois Tech guide to the role draws the line cleanly. A regular developer "focuses on one capability across many customers". An FDE "focuses on one customer across many capabilities".

The work runs in a loop that repeats on every engagement:

1. Discover. Sit with the people who do the work. The real problem rarely matches the statement of work.
2. Translate. Turn a messy workflow into a problem code can solve. "Claims staff retype numbers from PDFs into three systems" becomes an extraction pipeline and a review queue.
3. Build. Ship production code inside the customer's systems, not a prototype on sample data.
4. Feed back. Tell your product team what the product couldn't handle, so the next customer gets it built in.

The last step is easy to skip, and it matters. The FDE sees what breaks in the field before anyone else at the company does.

What does the work look like week to week?

The week mixes debugging, conversations and code, and the conversations often decide the code. Here is an illustrative week. Say you're placed with an insurer to get a claims assistant into production:

• Monday: the retrieval step times out against a document store nobody has documented. You read logs until you can make it fail on demand.
• Tuesday: the security team won't open a network route to the model endpoint. You explain what data leaves, where it goes and how it's logged.
• Wednesday: you ship the fix that Tuesday's meeting made possible, behind a flag, tested on a set of real claims.
• Thursday: you walk the customer's own engineers through the code, because they will run it after you leave.
• Friday: you write up the two things the product couldn't do and send them to your product team.

No two weeks match. The constant is ownership. You're accountable for whether the system runs, not for whether your advice was sound.

How is an FDE different from a solutions engineer or a consultant?

The difference is production code, and staying after the sale. A solutions engineer proves the product fits before the contract, with demos and proofs of concept. A consultant writes the recommendation. The FDE writes the code in the customer's systems and stays until it works.

Titles blur, so read the posting, not the title. We set the roles side by side, including what each interview tests, in FDE vs solutions engineer vs ML engineer.

Where did the forward deployed engineer title come from?

The title is most associated with Palantir, which calls the role Forward Deployed Software Engineer (FDSE) and built it for government and enterprise customers that needed working systems inside their own environments. Barry McCardel, who spent almost five years in the role there, describes how that culture worked. Palantir still hires under that name. Search the board for Palantir to see its current listings.

The title has since spread far beyond one company. The reason is the one Palantir had. A product that works in a demo still has to meet one customer's data, permissions and old systems. Someone has to do that work from the inside.

Who hires forward deployed engineers?

AI labs, software vendors, consulting firms and startups all hire FDEs, and the biggest posters on the FDEBench board are Google, Databricks and Salesforce. As of 2 Nov 2026, the board holds 1,453 roles from 799 companies, gathered since May 2026 and cleaned of duplicates.

The roles are spread across regions: 56% are in the US, 18% in Europe, 7% in India and 19% elsewhere. Browse the US roles, or start with the Core FDE family: listings titled as plain forward deployed engineer roles, without a solutions, data or AI qualifier.

The type of employer changes the job. An AI lab wants working fluency with models. A software vendor wants platform integration and data pipelines. A consulting firm or startup wants someone who can wire systems together and design how they fit, with model depth as a plus. How to read an FDE job posting shows how to tell which one you're looking at.

The skills postings tag most often are Python, AWS and Azure. Python alone appears on 33% of roles. That's the stack you'll be deploying into, so expect questions on it.

How much do forward deployed engineers make?

In the US, the median of posted pay-range midpoints is $175k for mid-level FDE roles and $208k for senior ones. Entry-level roles sit at $140k. Those figures come from the 462 US listings on the board that publish a USD range.

Each level has a different number of listings behind it, and that changes how far you can trust each figure. The full table by level, with each sample size and how we calculate it, is in forward deployed engineer salary.

What does the forward deployed engineer interview test?

It tests the customer side as hard as the code. The prep page sums up one bootcamp student's eight AI interview loops. They came down to three things:

• Customers and stakeholders: how you handle people who aren't technical.
• Technical: debugging real code and logs, or building a feature from scratch.
• The final round: a pitch, with a proof of concept built on the company's product.

If you come from engineering, the prep page's advice is to expect the customer side to be tested hardest. So practice explaining a system to someone who holds the wrong picture of it: "A stakeholder believes the model ‘connects to the internet’ to answer questions with fresh information. Correct the picture and describe the actual data flow." That question has a full scoring rubric.

Then try two harder versions of the same skill. One is a customer who thinks replication across three availability zones keeps their data safe. The other is a deployment that ‘can reason about the problem but never actually does anything’, where you explain what's missing and what you'd build.

For the technical side, pick a design question that has a rubric, like a multi-tenant RAG assistant returning other customers' documents. The round-by-round guide is in the forward deployed engineer interview, and all 159 questions, 22 with a full rubric, are on the prep page.