FDE Hub logo: bright terminal prompt and forward arrow joined at a hub nodeFDEHUB.DEV

Learn · Curated voices

Voices from X: Forward Deployed Engineering

A curated reading list of high-signal posts, threads, and talks from practitioners who have done the job, built FDE teams, or studied the Palantir model closely — not a firehose of job ads.

Last reviewed · Third-party views with editorial notes · Syntheses: real FDE vs renamed services · FDE product loop

How this page is curated

X is full of FDE job posts and title inflation. A post belongs here if it does at least one of the following:

  • Distinguishes FDE from solutions engineering, consulting, or delivery
  • Comes from someone who has been an FDE, hired FDEs, or ran an FDE rotation
  • Treats field work as a product loop, not just onsite implementation
  • Gives a usable test for whether a company is building software or scaling services under a better title

We favor judgment over engagement. Likes are a weak filter. “Hiring FDEs” with no definition is excluded.

Definitions and original doctrine

What the role was invented to do, in the words of people close to the Palantir model.

  • Lenny Rachitsky × Nabeel Qureshidefinitionpalantirinterview
    Forward-deployed engineering explained

    Nabeel spent nearly eight years as a Palantir FDE. The line worth keeping: the job is not to deploy software and not to sell software — it is to solve the problem. Lenny’s interview remains the best long-form entry point for why the model also produces strong product people.

    Open on X · Full conversation

  • Lenny Rachitskydefinitionsolutions-engineereconomics
    FDE vs solutions engineer

    A clean 80/20 cut: in the Palantir sense, most of what an FDE builds is custom for one large customer, with intent to abstract later. Solutions engineers mostly integrate existing software with small changes. The model only works economically on very large contracts.

    Open on X

  • Palantirpalantiragentsplatform
    Palantir on the AI Forward Deployed Engineer

    Official framing of an agentic system that can build transformations, edit an ontology, write functions, and stand up applications. Useful as a signal that “FDE” is now both a human role and a product metaphor.

    Open on X

  • Ryan Singerdefinitionhistoryproduct
    Problem-shaping, not ticket-taking

    An early, widely cited post treating FDE as engineering that owns problem definition. Quotes Palantir’s Ted Mabrey: the company thought of the FDE model as one of its greatest secrets.

    Open on X

  • Hobsondefinitionaccountability
    FDE is not consulting, solutions, or delivery

    Consulting sells judgment and leaves. Solutions sells a design and a POC. Delivery sells a statement of work. FDE sells an outcome inside the customer’s production system — if it fails there, it is on you.

    Open on X

Product loop vs renamed services

The load-bearing test: after the deployment, did the product get better for the next customer?

  • Verne Field Notesproduct-loophandoffownership
    It ends when another team can run what was built

    A clean endpoint test missing from most FDE job posts: discovery starts inside an operation, and the work is not done at release — it is done when a named customer team can maintain the operating loop. Embedding earns its cost only when exceptions live in people’s heads and requirements cannot be learned from a brief; otherwise configure a product and stop.

    Open on X

  • FDE Hub (Mandic)product-loophandoffbuyer
    The move-out clause — and “FDE washing” from the buyer’s chair

    Buyers now ask what happens after the outcome: name knowledge transfer, write decreasing FDE hours into the contract, and treat editable process rules (not a PDF handoff) as the real transfer artefact. Mandic pairs that with Gartner’s “FDE washing” label — engagements that never feed product and leave nobody inside who can evolve the system. Same test Verne’s ownership exit makes from the vendor side.

    Open on X

  • FDE Hub (Mandic)product-loopmarketdefinition
    State of FDE 2026: product function in name, services shape in the code

    Twenty-four anonymous practitioner answers (no medians published): reporting lines split five ways, most customer code stays bespoke rather than merging back, and the week splits cleanly along whether you write production code — discovery-heavy seats and build-heavy seats look like two jobs under one title. Mandic’s operator tests: where does the code end up, and who owns what shipped six months ago?

    Open on X

  • David Senra × Mati Staniszewskiproduct-loopelevenlabspalantir-alumni
    The piece most companies miss

    ElevenLabs CEO and former Palantir Deployment Strategist: FDEs sit on the product team, not GTM. They stretch the product for one customer, then bring learning back so the next customer inherits it. Many companies only copy the first half.

    Open on X

  • Yumproduct-loopservices-trapproject-frontline
    The 100-hour test

    After 100 hours of customer implementation, did you create reusable primitives — integrations, permissions, workflows, defaults, evals, deployment patterns — or just finish 100 hours of custom work? If the same problems are still solved by hand after the 5th customer, that is services wearing a product title.

    Open on X · Follow-on (Project Frontline)

  • Nabeel Qureshititle-inflationhiringwarning
    Many “FDE orgs” are sales engineering with a cooler name

    Practitioner warning: a large share of new “FDE functions” are ordinary sales-engineer or solutions-architect roles being rebranded. Ask for production ownership and a product write-back loop before you believe the title.

    Open on X

  • Gergely Orosztitle-inflationplatformcritique
    You cannot be forward deployed from outside the platform

    A useful negative definition. A major sell of FDE was contributing back to the platform after seeing the customer. Roles that only bolt a model onto an account, then optimize token spend, drop that loop — treat them as a different job.

    Open on X

  • First Round Reviewhiringfounderswhen-not-to
    Does your startup actually need FDEs?

    Panel of people who built or hired FDE teams. One tell the model is a bad fit: you already have a crisp vision of exactly how customers should use the product. Palantir’s early advantage was the opposite — FDEs discovered niche uses in the field.

    Open on X · First Round Review essay

  • Richard Seroter → Latent Space / Vinoo Ganeshproduct-looplongform
    If you are not an extension of product/engineering, you are a renamed sales engineer

    Seroter’s one-line filter attached to Vinoo Ganesh’s Latent Space essay on doing the job right. Ganesh led Spark at Palantir and built Project Frontline — the rotation that turned software engineers into FDEs.

    Open on X · Latent Space essay

Talent, prestige, and the market

Why the title is spreading, what that does to hiring, and where the public conversation concentrates.

  • FDE Hub (Mandic)talentmarketdefinition
    Industrial FDE vs original Palantir FDE

    Hyperscaler hiring is flooding the market with short-cycle “industrial” FDE pods. That is not the same job as deep embed with product write-back. The screen question: which FDE are you hiring for — and which one do you want to become?

    Open on X

  • FDE Hub (Mandic)talenthiringtitle-inflation
    The recruiter template is wrong — read the title twice

    Thirty inbound FDE messages as a dataset: the copied template (Python + SQL + pipelines + customer-facing) describes a data engineer who attends meetings, not discovery, scoping, or outcome ownership. Practical test: read once for the title the market rewards, once for the work you would actually do — in this sample they agreed less than half the time. Headline hyperscaler demand and agency-intermediated inbox demand are different markets.

    Open on X

  • FDE Hub × Kanav Bhatnagartalenthiringstartup
    Two archetypes: AI-native FDE vs mature SaaS FDE

    Actively (AI-native) vs Rippling (AI-forward SaaS): one day is building the AI product and pre-sales POCs; the other fills gaps on a mature platform after sales already won. Agree P&L metrics before the pilot, go narrow end-to-end, and know that multi-customer economics usually force breadth over Palantir-style single-account depth.

    Open on X

  • Aaron Levie + Gergely Orosztalentcareersai-agents
    Make FDE visible to students

    Levie’s pitch to career services: CS depth, systems thinking, business acumen, and fluency with coding agents / MCP. Orosz adds that FDE may replace the consultant / solutions-architect on-ramp that absorbed many new grads from 2010–2022.

    Open on X · Gergely’s commentary

  • Jaya Guptatalentprestigehistory
    The FDE model is about talent status, not just deployment

    McKinsey made client service prestigious for business generalists. Palantir made embedded deployment prestigious for technical generalists. The AI-era question: who makes implementation feel like cutting-edge work?

    Open on X

  • Trace Cohenenterprise-adoptionone-liner
    Consultant 2.0

    Compressed adoption problem: Palantir had to send A+ talent into messy environments because that is the only way A+ software gets used.

    Open on X

  • swyx / AI Engineer World’s Fairtalkssurveyhiring
    A field survey in one conference track

    The most complete public snapshot of the discipline in one place: FDE leaders from Cursor, Sierra, Ramp, Cognition, Factory, Anthropic / ex-Palantir, Kepler, and more. Start with Kevin Bai’s FDE 101 for a single talk.

    Open on X

Getting started notes

Practitioner roadmaps and field craft. Personal summaries — not official curricula.

  • YOUSSEF EL BEQQALexplainerbeginner
    What the gap actually is

    Most enterprise software fails because implementers do not understand the business and operators cannot implement. FDE sits in that gap — especially when every customer’s data, legacy stack, and definition of “correct” differs.

    Open on X

  • Ritesh Roushanroadmapskillsbeginner
    A practical skill ladder

    Splits the job into engineering depth, messy integrations, customer translation, and the AI deployment layer (RAG, agents, evals, on-prem, access control). Growth path: SDE → integration → workstream owner → senior FDE who feeds product.

    Open on X

  • WebDevCodyworkflowagentsfield-craft
    The tight demo loop

    A current working style: sit with the client, dump org context into a knowledge base, have a coding agent produce tomorrow’s demo, repeat. Incomplete as a full FDE definition — useful as a picture of how field cycles are speeding up.

    Open on X

Accounts to follow

Not exhaustive. These accounts reliably add signal instead of restating the job description.

  • @nabeelquPalantir FDE originalism; allergic to title inflation
  • @lennysanBest interviews on the role and the companies copying it
  • @swyxConference scan of the whole FDE field
  • @GergelyOroszHiring market + skepticism about fake FDE roles
  • @firstroundWhen startups should and should not copy the model
  • @davidsenraFounder interviews; Mati / ElevenLabs segment is essential
  • @matiProduct-team FDE, not GTM FDE
  • @PalantirTechOfficial definition and AI-FDE product direction
  • @natalie_meurerSierra; title drift and agent engineering
  • @levieDemand-side view from enterprise software
  • @JayaGup10Talent prestige and why the role is spreading

Related long-form (off X)

Frequently asked questions

What is Voices from X on FDE Hub?
A curated reading list of high-signal X posts, threads, and talks from people who have done Forward Deployed Engineering, built FDE teams, or studied the Palantir model closely. It is not a firehose of job posts.
How do you decide which FDE posts to include?
A post belongs here if it distinguishes FDE from solutions engineering, consulting, or delivery; comes from someone who has been an FDE or hired FDEs; treats field work as a product loop; or gives a usable test for software vs services under a better title.
Are these posts official FDE Hub doctrine?
No. They are third-party views with editorial notes. Prefer durable linked sources and always attribute the author.