The article explains how Forward Deployed Engineer and AI Engineer roles, originally distinct in focus and metrics, are converging in skills but remain differentiated by where they work and how their performance is measured.
Aug 12, 2026
·
4 min read
"Forward Deployed Engineer" and "AI Engineer" now appear on job posts describing nearly identical work — but one starts at the customer and works toward the model, and the other starts at the model and works toward the customer. Here's how to tell which one a post is actually offering.
How two of the most-posted titles in AI ended up meaning almost the same thing — and the single question that separates them.
The problem: the titles stopped being informative
Job titles are compression. You read one and you know roughly what the work is, who you'll sit with, and what you'll be judged on. For most of software this holds up — "backend engineer" tells you something, "SRE" tells you something.
These two don't. Read enough postings and you'll find Forward Deployed Engineer roles that are pure product work, and AI Engineer roles that are four days a week in a customer's office. The title above the description has stopped predicting the description.
A title that means two things isn't compression. It's a coin flip.
What's needed is a test that works on the job description itself, not on the label stapled to the top of it.
What each one actually is
The Forward Deployed Engineer is an engineer who goes to the customer. They sit in the customer's building, learn a workflow nobody documented, and build on top of their own company's platform until it does that customer's specific job. The pattern is generally credited to Palantir, which paired the FDE with a non-engineering counterpart, the Deployment Strategist. It has since been picked up across applied AI — OpenAI, Anthropic, Sierra, Harvey, Decagon, Glean.
The FDE isn't selling, and isn't quite building the product either. They close the distance between a general system and one organization's particular mess.
The AI Engineer is an engineer who builds products on top of foundation models through APIs rather than training models themselves. The term was popularized by swyx (Shawn Wang) in an essay called "The Rise of the AI Engineer," which argued that models had become capable enough to support a new engineering discipline sitting between the ML engineer and the product engineer. The AI Engineer Summit and World's Fair grew out of it.
The AI Engineer treats the model as a component. Prompts, evals, retrieval, tool use, latency, cost.
Side by side
| Forward Deployed Engineer | AI Engineer | |
|---|---|---|
| Starts from | One customer's workflow | The model's capabilities |
| Sits with | Customer ops teams, solutions, sales | Product, design, platform |
| Ships | A working deployment for one account | A feature for every account |
| "Done" means | The customer's process runs | The eval passes and it's in prod |
| Core skill | Reading an organization | Reading a model's failure modes |
| Fails by | Building something unrepeatable | Building something nobody needed |
| Measured on | Accounts live, renewals, expansion | Shipped features, quality, latency, cost |
That last row is the one that actually decides your day. Everything above it is negotiable at any given company; the number you're graded on is not. Two roles can share every skill on the list and still be completely different jobs because one is measured in accounts and the other in releases.
Why they're converging anyway
Two forces pushed these roles toward each other.
First, the bottleneck moved. When models were the hard part, the scarce skill was building them. Once general capability got cheap, the hard part became the last mile — the gap between a system that works in a demo and one that survives contact with a real company's data, permissions, and habits.
Second, every applied AI company hit the same wall at roughly the same time. The demo works. The deployment doesn't. Not because the model is bad, but because nobody on the team understood the customer's edge cases well enough to encode them.
So FDE postings started asking for evals, retrieval, and prompt work. AI Engineer postings started asking for customer calls and on-site time. Both were reaching for the same missing person from opposite sides.
The honest take
The strongest argument against everything above: they aren't converging, and treating them as one role is exactly how people end up in a job they didn't want.
The FDE is structurally a go-to-market role wearing an engineering title. It frequently reports into a revenue or solutions organization, not into engineering. That changes the comp shape, the promotion ladder, the definition of a good quarter, and who decides what you work on next. An engineer who takes an FDE role expecting product work usually isn't disappointed by the code — they're disappointed by the org chart.
The convergence is real at the level of skills. It is much weaker at the level of incentives. Skills are what you use; incentives are what happens to you.
Trade-offs
- Leverage. The AI Engineer's work compounds — one feature serves every customer. The FDE's work often doesn't leave the account it was built for. Some companies fix this by routing FDE learnings back into the product; most say they do and don't.
- Optionality. "AI Engineer" reads cleanly to any future employer. "Forward Deployed Engineer" needs a paragraph of explanation outside the handful of companies that use it, and that paragraph is doing a lot of work in a resume screen.
- Proximity. FDEs learn what customers actually do, which is the most undervalued input in product work and nearly impossible to acquire from a desk. It usually costs travel and a calendar you don't control.
- Depth. AI Engineers go deeper on the model. FDEs go deeper on the domain. Both are real specializations; neither is a fallback for failing at the other.
Verdict
If you're staring at a job post and can't tell which role it is, ignore the title and read for two things: whose calendar you're on, and what number you're judged by. If the answer is a customer's calendar and a count of live accounts, it's an FDE role no matter what the header says. If it's your team's calendar and a shipped roadmap, it's an AI Engineer role even if the post says "forward deployed."
Ask it in the interview if the post won't tell you. A company that can't answer cleanly hasn't decided either — which is its own useful signal.
Same models. Different rooms. Different scoreboards.