Module 01 · 8 min

What Is an FDE?

Everyone agrees the role is hot. Almost nobody agrees what it is. This module gives you a definition precise enough to argue with — and a test you can run against any real job posting.

The normal shape of software engineering One capability you build it once Customer Customer Customer Customer …thousands more The forward deployed shape — Palantir calls it “Delta” Data integration Custom app Agent + evals Security model Adoption plan ONE customer or one mission The distinction is accountability, not travel Discovery Architecture Build Production Adoption FDE owns all of it A demo is not the finish line. A system people use is. Why the AI era made this urgent 🎲 Probabilistic Models degrade quietly. Somebody must build evaluations. 🔐 Proprietary data The value is locked in systems nobody outside the building can see. and one more… The third pressure: field learning outruns the roadmap Product team plans the roadmap The field reality, at speed ships the FDE carries this signal home Without that return arrow, you have a consultancy. So: which one are you looking at? GENUINE FDE Owns production repos Discovery → adoption Field work reaches product Measured on outcomes THE TITLE ONLY Configures and demos Hands off before prod Code dies in a branch Measured on utilisation Module 02 turns this into a test you can run.

A definition precise enough to argue with

A Forward Deployed Engineer is a software, data, machine-learning or systems engineer who is assigned close to a customer or mission environment and is accountable for turning a vendor's product or technical capability into a working production outcome.

Read that twice, because every clause is load-bearing. Engineer — not adviser. Close to a customer — not to a codebase. Accountable — not contributing. Working production outcome — not a proof of concept, not a demo, not a recommendation deck.

This is consistent with the actual language in official postings. OpenAI measures FDE success through production adoption, workflow impact and evaluation-driven feedback. Google describes the position as an embedded builder responsible for the connective tissue between AI products and customer infrastructure — and pointedly notes the role is not advisory. Scale AI expects daily customer interaction, full-stack delivery and product-roadmap influence.

The instrument

The eight tests

A role should satisfy most of these to count as a substantive FDE position. Tick the ones that apply to a job you're actually looking at — the verdict updates as you go.

Six or more is a substantive role. Four to five means it's a hybrid — go and check the decision rights before you accept. Three or fewer and you're looking at an adjacent role wearing the title.

Where it came from

Origins and evolution

Palantir popularised the contemporary model. Internally it uses three names: Delta for forward-deployed software engineering, Dev for core software engineering, and Echo for the deployment-strategy function.

Their own comparison is the cleanest one-liner in the whole field: a conventional developer builds capabilities that can serve many customers, while a Delta enables many capabilities for a particular customer or mission.

Deployment Strategists tend to lead problem framing and operational adoption. Palantir itself notes that the boundaries between strategist, product manager and engineer blur in practice — which is honest, and worth remembering whenever someone tells you the categories are crisp.

Evolution of the model

2010s
Palantir popularises Delta / FDSE Engineers embed with institutions and operational users. The model is unusual enough to be a curiosity.
Early 2020s
Adjacent models spread Cloud, data-platform and robotics companies adopt resident architects and field engineers with broader coding ownership.
2024
The pilot graveyard Enterprise generative-AI pilots expose integration and adoption gaps. Applied-AI and deployment-engineering titles proliferate.
2025
Demand accelerates Frontier labs, startups and consultancies all hire. Reported posting growth approaches eightfold during parts of the year.
2026
Formalisation and industrialisation Major vendors formalise multi-level FDE ladders. Global systems integrators begin operating the model at scale — TCS alone plans 5,900 to 8,900 FDEs.

What it is not

Eight roles get confused with FDE. Here's the one-line version — Module 02 turns it into a game you can lose.

Adjacent roleTypical distinction from a strong FDE
Solutions engineerMore pre-sales, demonstration and technical validation oriented
Solutions architectConcentrates on architecture and recommendations rather than direct implementation
Implementation engineerExecutes a defined configuration or migration with less product influence
Professional-services consultantScoped by statements of work, utilisation and billable delivery
Customer-success engineerSupports adoption, health and retention after deployment
Site-reliability engineerOwns reliability of a platform, not the customer's complete business outcome
Field engineerDeploys physical or industrial systems — an FDE equivalent when substantial software development is involved
Sales engineerSupports revenue acquisition; generally does not own long-lived production code
⚠️ The boundaries are porous — deliberately so

Databricks' FDE-labelled opening is explicitly billable and housed in Professional Services. Google's posting stresses the FDE must move prototypes to production-grade agentic workflows. Same title, two genuinely different jobs. The title alone cannot determine the operating model — which is why you need the eight tests.

Check yourself

Module 01 quiz