Where the roles are listed, and why the listings mislead

There is no shortage of listings and there is a shortage of accurate ones. Third-party indexes in late 2026 carry a few hundred to roughly a thousand roles between them, and that number is unreliable in both directions: it counts the same job posted by three recruiters, and it misses roles that are the same work under a name no index searches for.

Three employer shapes

Model labs. The role was invented here and the compensation is the highest. The work is closer to product engineering with a travelling component; you are usually one of a small number deployed at a flagship customer, and the feedback you carry back goes directly into the model or the product surface. Few seats, high bar, and the interview is heavy on building.

Deployment shops and integrators. The company's entire business is the deployment. You will do more of the job's unglamorous half — scoping, stakeholder management, handover — and you will see many customers rather than one deep one. This is the best environment for volume of reps, and the worst for anyone who wants to touch model internals.

Vertical SaaS and large enterprises. The title usually is not FDE. It is solutions architect, implementation engineer, AI application engineer, customer engineer, or technical account manager. These are the roles an index misses, and half of them are genuinely the job while the other half are pre-sales with an engineering vocabulary. The questions in is an FDE a pre-sales role are how you tell.

Reading a listing

Four signals, in order of reliability.

1. What is the success measure? "Drive adoption" is a verb with no number. A listing that names a metric — time to first value, activation rate, deployment count — was written by someone who has run the work.
2. Where does the role sit? Under Engineering or Product: probably the real job. Under Sales or Revenue: probably quota-adjacent. Under Customer Success: usually support with travel.
3. What is the travel expectation, stated in numbers? "Up to 50%" is a real answer; "some travel" is not. A deployment role with no travel is either remote-first tooling or a role that has been quietly hollowed out.
4. Is there a named handover artefact? Listings almost never mention one, which is why the ones that do are worth taking seriously.

Two things to discount. "Own the customer relationship end to end" often means there is no account team, which means you are also the escalation path at 11pm. And "wear many hats" in a company of forty means the hat is sales.

Japan, and why it lists separately

The Japanese market has its own job board for this role, which tells you something about how the occupation is forming: it is being imported as a distinct job family rather than grown out of existing consulting titles. Listings there lean toward enterprise AI adoption projects and often specify Japanese-language customer work, which changes the candidate pool more than the technical bar does.

The channel that beats the boards

Most of the roles that fit well are filled before they are listed, for the ordinary reason: a deployment is a trust purchase, and a referral from someone the customer already trusts is worth more than a pipeline of applicants. The practical consequence is that the board is for market intelligence — who is hiring, at what level, with what vocabulary — and the actual entry usually comes from a conversation.

That conversation is easier to start than it sounds. A public artefact helps: a written teardown of a deployment problem in the employer's own vertical, posted where their engineers read. It is the same evidence that a portfolio would carry into an interview, produced before anyone asked.

What the numbers say, cautiously

Across the public indexes the distribution is lopsided: a small number of labs and funded deployment companies account for a large share of listings, and most of the rest are single-role postings from companies testing whether the function is worth a team. Treat any total as a floor, because the untitled roles — implementation engineer, AI application engineer — are not in these counts at all.

What to do with that: apply where the deployment is the business, because those teams have a method and will teach it; and treat the titled index as a starting map rather than a market. How the pay bands differ across those three shapes is in the ladder and pay, and what the interview tests is in what an FDE interview actually tests.