FDE Jobs Are Real. The Taxonomy Is the Hard Part

What the current open-job sample says

We queried the production index for titles containing the Forward Deployed Engineer phrase or the FDE abbreviation, then constrained results by employer-reported posting date. The corpus snapshot contained 2,679,118 open jobs from 60,835 companies across 38 ATS source families. Of the records that met the title query, 69 had a posting date in the prior seven days, 231 in the prior 14 days, and 443 in the prior 30 days. The 60-day cut contained 543.

Raw counts are hard to interpret when the denominator changes, so we compared each count with all currently open postings carrying a date inside the same window. FDE-labeled roles represented 0.074% of the seven-day sample, 0.076% of the 14-day sample, 0.077% of the 30-day sample, and 0.068% of the 60-day sample. That is a modestly higher share in the shorter cuts, but the 7-, 14-, and 30-day values are effectively a plateau—not evidence of an accelerating curve.

This distinction matters because nested windows are not a historical time series. The 30-day set contains the seven-day set, and all four cuts include only jobs still open on the snapshot date. A role that was posted and closed quickly can disappear from every cut. These numbers describe the composition of the current open index by posting-age threshold. They do not measure hires, applications, headcount, or net job creation.

One acronym now covers several different jobs

The sample titles reveal why exact-title search is brittle. They include Principal Forward Deployed Engineer, Manager of Forward Deployed Engineering, AI Engineer — FDE, AI Process Forward Deployed Engineer, Customer Architect (Forward Deployed Engineer), and implementation-oriented variants. Some are deeply technical individual-contributor roles. Some combine solution architecture with customer delivery. Others look closer to technical consulting, deployment, or post-sales engineering.

A job seeker who searches only for software engineer may miss relevant FDE work. A search for solutions engineer may return the role, but bury it under pre-sales positions. A taxonomy that maps every FDE to engineering loses the customer-facing component; one that maps every FDE to sales engineering loses the production-building component. The label is stable enough to deserve recognition and unstable enough to resist a single flat category.

This is a common pattern when software companies operationalize a new delivery model. The title begins as company-specific language, spreads to adjacent firms, and acquires local variants before occupational taxonomies catch up. During that interval, product teams face a choice: wait for a standard occupation code to settle, or represent the emerging role explicitly with evidence and uncertainty.

Treat FDE as a title family, not a new universal occupation

For retrieval, start with an explicit alias family: forward deployed engineer, forward-deployed engineer, forward deployed engineering, and FDE when the surrounding text supports the expansion. Do not match the acronym everywhere. Three letters are cheap, and false positives compound quickly across millions of records. Phrase matches can carry high confidence; acronym-only matches should require corroborating terms such as deployment, customer, implementation, solution, AI, or engineer.

Next, separate the source title from the normalized representation. Preserve the exact employer title for display and audit. Add multiple facets rather than forcing a single label: engineering as the likely career family; customer deployment as a work activity; solution architecture, implementation, and production integration as potential skill clusters; individual contributor or manager as seniority. A search product can then retrieve FDE roles from several reasonable entry points without rewriting what the employer published.

Finally, allow the title family to change. Store the taxonomy version and the evidence used for classification. Review a stratified sample by source, geography, and seniority. If a growing share of matched records is sales-led rather than build-led, the retrieval weights should move. The goal is not to declare the eternal definition of FDE. It is to make today’s results useful while keeping tomorrow’s correction possible.

• Keep source_title immutable and display it to users.
• Store matched aliases and classification confidence for audit.
• Use multi-label activities and skills instead of one forced occupation.
• Sample acronym-only matches separately because they carry the highest false-positive risk.
• Version the title family so alerts and analytics can be reproduced after a definition changes.

What this changes for a job board

A niche AI job board should expose FDE as a navigable landing page only when it can maintain enough active inventory and explain inclusion. The page should include related labels, representative activities, and the last refresh time. It should not claim that every result is equivalent. Filters for seniority, customer exposure, travel, programming requirements, and deployment environment may be more useful than a generic department filter.

For alerts, title-family membership must be evaluated when a record is first observed and whenever the title or description changes. Otherwise a renamed role can remain in the wrong alert stream. For matching products, the hybrid nature of the role argues for evidence-weighted matching: a candidate with software delivery plus customer implementation experience may be a stronger fit than someone whose profile merely contains the exact title.

For workforce-intelligence buyers, count changes should be reported with the query definition, denominator, coverage window, and revision policy. A chart labeled ‘FDE demand’ without those details invites a stronger conclusion than the data can carry. The interesting commercial capability is not producing one dramatic number. It is detecting a new title family early, measuring it consistently, and letting a buyer inspect the underlying postings.

Limitations and the next measurement

This analysis uses current open postings, not a frozen archive of everything published in each period. The posted_since filter operates on date_posted, an employer- or source-provided field that is absent from a material share of the corpus. The title query can miss roles whose descriptions describe forward deployment without using the phrase, and it can include acronym collisions. Total estimates are appropriate for directional product analysis, not official labor statistics.

A proper trend study needs weekly immutable snapshots, a reviewed title classifier, and a stable denominator. It should distinguish new postings, active stock, expirations, and re-openings. It should also compare the title family with adjacent roles such as solutions engineer, implementation engineer, customer engineer, and technical consultant. Only then can we say whether the occupation is gaining share, being renamed, or absorbing work that previously sat elsewhere.

The current result is still valuable. Four hundred and forty-three recent open postings are enough to show that FDE is not a one-company curiosity. The flat near-term share is enough to reject the easiest hype headline. For B2B job products, that combination is the point: production data should help decide both what to build and what not to claim.