Deployment RunReturn to your game

Deployment Run: the FDE learning game

You’re the engineer. The customer needs it to work.

Preparing your first customer mission…

Learn the work.
Build the evidence.

A practical path into forward deployed engineering.
Take one customer problem from discovery to handoff.

Your progress0%0 of 29 milestones complete

Your next milestone

Write and debug a useful program

Start with a small tool you can run, test, and explain.

Start this milestone

Adjust your learning planNew to engineering

Choose your starting experience

Read the two sources together12 sessions, with recall questions and artifacts Try the field exercisesMake a decision, inspect the outcome, build your project

Before you start

Do I need a degree or a particular job title?

Requirements depend on the employer and level. A graduate opening is different from an experienced AI deployment role. Read the exact requisition, then show relevant software, domain, and customer experience. A course certificate alone does not establish those abilities. Compare entry routes.

How long will this take?

The estimate above budgets time for this guided project and changes with your background and study hours. It is not a hiring deadline or the length of every linked course. New programmers may need substantially more practice. Experienced engineers can demonstrate a milestone and move on without repeating material.

What should I learn first: AI, data, or software engineering?

Start with the ability to write, test, and debug a small program. Add SQL and APIs, then learn to understand a customer workflow. AI becomes useful when you can integrate it into a system and evaluate whether it improves that workflow. You do not need to train a foundation model to practise this path.

How do I get customer experience before my first FDE role?

Try an internal tool at work, an open-source integration, or a tightly scoped volunteer project with a willing organisation. Agree on permissions and an operating owner. If you use a fictional customer, label the project as a simulation and ask a peer to play the user. Never invent adoption or business-impact figures.

What should I pay for?

The linked curriculum is available to read without buying a bootcamp. Optional credentials, hosted services, and model usage can cost money. Start with local data and mocked model calls, then set a small usage budget before connecting a paid service. Buy support or feedback only when you know what gap it will close.

This is an independent study plan, not an accredited course or a hiring guarantee. Milestones record your own assessment. Download the complete curriculum.

Practise the whole engagement.

Make decisions in the Northstar simulation, then apply the same steps to your project.

Open the paired reading companion

FDE workflow atlas

See the whole job. Work one stage at a time.

Follow an engagement from first observation to durable ownership. Open every stage to study its process, inspect a source-backed public example, practise a decision, and turn your feedback into a next step.

The interactive atlas needs JavaScript. Read the six-stage workflow and public cases; the learning path and complete curriculum remain available above.

Build throughout the path

One customer problem.
A complete deployment.

Carry the same project from a discovery conversation to a system someone else can operate. Your portfolio grows with each stage.

Open the full production checklist

Read with a job to do.

Benmore’s Forward Deployed and Panaversity’s AI Agent Factory, connected through twelve practical reading sessions.

Download reading plan

Read the linked section, close it, answer the recall question, then build the artifact. Reading progress is separate from demonstrated milestones.

Two readings. One delivery practice.

Use Benmore for discovery and engagement judgment, and Agent Factory for specifying and operating AI workers. Apply both to the same customer problem.

Read with a concrete artifact to build
Reading	Start with	Build next
Forward Deployed · Richard Buehling	Chapters 4–8: discovery and definition. Chapter 13: handoff.	An evidence ledger, a conditional scope, and an ownership drill.
The AI Agent Factory · Panaversity	Orientation, spec-driven development, governed knowledge, and evaluation.	A testable specification, source register, and bounded worker.

Download the field workbook · Chapter guide and source limits · Review the delivery gates

The workbook is an independent synthesis. Benmore’s original PDF is supplied separately; the website hosts its citation and reading guide.

Understand forward deployed engineering

Market evidenceAnnouncements and observed roles

Forward deployed engineering combines customer discovery with hands-on delivery and responsibility for a working system. Employers use the title differently. Read the scope, success measures, and operating constraints before treating two roles as equivalent.

What changed in 2026

Company announcements establish investment and intent; they do not measure the whole FDE labour market.
Date	Development	Implication for candidates
23 February	OpenAI announced multi-year Frontier Alliances with BCG, McKinsey, Accenture, and Capgemini, working alongside its FDE team. OpenAI	Look at delivery partners as well as model developers. Confirm the engineering ownership of each opening.
4 May	Anthropic announced an AI services company with Blackstone, Hellman & Friedman, and Goldman Sachs. Its announcement names no capital amount. Anthropic	Mid-sized enterprises are also a target for embedded AI engineering.
11 May	OpenAI launched DeployCo, majority-owned and controlled by OpenAI, with more than $4B in initial investment and 19 partners. OpenAI	DeployCo has its own hiring page; distinguish the employing company from OpenAI itself.
July	AWS described a $1B investment in its FDE organisation and an emphasis on customer self-sufficiency after delivery. AWS	A good handoff and customer capability are deliverables, alongside working software.

A correction to the DeployCo story

OpenAI’s launch announcement says its agreement to acquire Tomoro would add approximately 150 Forward Deployed Engineers and Deployment Specialists. It described the transaction as subject to closing conditions. That supersedes this guide’s earlier account of 95 engineers and five investors. The announcement does not substantiate a valuation, guaranteed investor return, or completed closing; those are not asserted here. OpenAI, 11 May 2026

What the job-market numbers can tell you

Henley Wing Chiu’s analysis used 1,000 postings and separately examined 100 FDE profiles. Its classification was 60% builder roles, 30% sales-engineering roles, and 10% internal-tool roles. It is useful evidence of title variation, with a last update of 25 January 2026; it is not a current census or a representative estimate of your hiring odds. Henley Wing Chiu · Bloomberry

Job-board totals also depend on whether managers, strategists, internships, security roles, and duplicate locations count. This guide does not calculate a global vacancy count, growth rate, or placement probability. A live advert is evidence of recruiting intent, not proof that a seat is unfilled.

What work does this FDE posting actually describe?

Nine functional testsPaste a posting, or tick by hand

Use the responsibilities to distinguish production ownership, sales support, account management, and delivery services. This local keyword checker is a conversation aid, not a validated classifier. Review its matches and ask the hiring team about anything the posting leaves out.

Paste the posting

The text stays in your browser. Reading a posting replaces the checks with keyword matches; adjust them by hand afterwards. Score shares are not probabilities.

Job description text

Nothing pasted yet

Treat the score as a prompt for questions

OTE, a sales reporting line, or work before a contract is signed does not by itself disqualify an engineering role. Databricks explicitly describes its AI FDE team as delivering professional services while owning production rollouts. Ask who writes the code, who maintains it, what success means, and how discoveries influence the product. Databricks

Role shape

Nothing ticked yet — tick the statements that match the posting.

The functional test that matters: are you embedded with one customer, writing production code in their environment, and accountable for whether the deployment delivers? If yes, it is FDE work whatever the badge says.

Adjacent roles and overlapping responsibilities

Six role patternsAn editorial comparison

Titles and reporting lines vary. Compare the recurring work and accountability below, then ask the hiring team which pattern describes this particular role.

Engineer, strategist, architect: ask about the actual split

Palantir’s postings distinguish engineering roles on Delta from Deployment Strategists on Echo. A team label is not a universal industry definition. The London FDSE advert emphasises end-to-end delivery; the role also includes customer relationships, architecture, data, and applications. Palantir

OpenAI’s FDE role owns deployment delivery, while its separate FDSWE posting emphasises reusable software and abstractions across engagements. Both involve customer work. OpenAI OpenAI

Professional services can include FDE work

Databricks describes its AI FDE team as a professional-services function that builds and productionises AI applications and influences product priorities. Billing model, department name, and engineering responsibility are different dimensions. A useful distinction is how much technical ownership, customer accountability, and reusable product learning the position gives you. Databricks

Choose a field of practice

Four flavoursEight skill axes each

Choose a domain to see an editorial weighting of the skills it often needs. These are preparation guides, not measured percentages of a working week. Company examples describe overlapping work; current openings and locations are in the dated directory.

Your skill map

Seventy-one skillsFive stages · your marks are saved

Explore specific skills in more depth, or use the guided learning path to decide what comes next. Open a skill for exercises and proof of competency. Use its status control to track your practice.

Progress

Hours you can study each week

Foundation coverage

0% of the core skills in Foundation and Job-ready

By branch

A capstone that strengthens your application

One flagship customer-facing system, built end to end and written up in public: discovery notes showing you found the real problem, an architecture decision record, integration reliability under a hostile API, an eval harness with documented failure modes, a live deployment with monitoring, and a case study written for a business reader. Then two mock interviews — one coding and debugging, one customer solution design — and a CV rewritten around ownership and measured outcomes rather than technologies used.

A portfolio project that demonstrates deployment ownership

Carry an evidence ledger and delivery specification into the build. Track prototype gaps, version the knowledge your workflow uses, and finish with an ownership drill and support boundary. Use the nine-part field workbook.

Choose a bounded workflow such as support-ticket routing, invoice reconciliation, or internal knowledge search. Use synthetic or openly licensed data. Work with a real user when possible; label a simulation honestly. The following is an editorial project brief, not an employer’s take-home or a promise of interview success.

Deliverable	Evidence to show	Failure to exercise
Discovery brief	User, workflow, current baseline, decision owner, success measure, and explicit exclusions.	The requested feature does not solve the user’s actual bottleneck.
Working integration	A small UI or API, persistent state, documented schemas, permissions, and a reproducible setup.	Expired credentials, missing records, schema change, duplicate events.
Evaluation suite	Representative cases, held-out examples, deterministic checks where possible, and human review of ambiguous cases.	Confident wrong answers, invalid tool arguments, unsafe actions, and regression after a model change.
Operating plan	Latency and cost per completed task, retry limits, logs with sensitive fields removed, alerts, and a rollback.	Timeouts, dependency outages, overloaded queues, and exhausted budgets.
Adoption and handoff	User feedback, an operating runbook, a named maintenance owner, and a short demo of recovery.	The customer cannot operate the system without its original author.
Reusable improvement	A documented component or product issue extracted from the engagement.	A second customer needs a different schema or permission model.

Measure the outcome, then inspect how it happened

Anthropic’s agent-evaluation guidance separates a task, repeated trials, the execution trace, and the final environment state. A statement that a ticket was updated is not proof that its database record changed. Combine executable checks, calibrated model graders, and human review; examine valid alternative solutions before calling them failures. Anthropic

For your project, report task completion on a held-out set, the error categories, repeated-trial variability, p95 latency, cost per completed task, and the rate of human escalation. Publish the denominator and sample selection. These are suggested project measures, not an industry hiring benchmark.

Choose the smallest architecture that passes the test

Start with a deterministic workflow or a single model call when it is sufficient. Add retrieval, tools, or an agent only when the task needs them. More autonomy can increase latency and cost. Anthropic

Indexed retrieval, live search, and a hybrid approach have different freshness and latency trade-offs. Context engineering does not mean that vector search is obsolete. Anthropic

For MCP integrations, record the protocol revision and SDK, exercise authentication and denied access, and test against the customer’s actual client. The July 2026 revision changes the protocol core; documented SDK compatibility with older revisions means migration cannot be reduced to deleting every old session implementation. MCP · David Soria Parra and Den Delimarsky Model Context Protocol

Prepare for the interview

Four employersPolicy checked 14 September 2026

Use official process descriptions where they exist, and practise the work the job requires. The exercises below are original preparation prompts. They are not a verified question bank or a guarantee of your interview sequence.

Get the tool rules for every round

Practise three additional prompts: challenge an unsupported claim in a customer brief; show how a changed requirement updates the specification and tests; explain why a receiver’s failed rollback drill blocks handoff. Prepare a bounded delivery proposal.

OpenAI’s public guide now explicitly says that AI permissions vary by interview and are specified in preparation materials. The earlier statement here that it publishes no policy was wrong. OpenAI

Anthropic allows AI for preparation and refining your own application, while assessments and live interviews are unaided unless the instructions say otherwise. Anthropic

Sierra’s published onsite uses an AI-assisted build. That permission should not be assumed to cover every other assessment or every employer. Sierra · Vijay Iyengar, Arya Asemanfar, Angie Wang

Three ways a working day can look

Illustrative schedulesTeam and engagement dependent

These are editorial examples of an on-site day, a build day, and launch support. They are not measured time-use data. A team’s staffing, travel, escalation load, and customer maturity can change the schedule substantially.

Questions about sustainable working conditions

Ask how many accounts you would carry, who covers urgent incidents, how travel is scheduled, and how the team protects time to build. Discuss what happens when a customer is blocked on access or when an executive sponsor leaves. Ask for a recent example of a deployment that missed its target and how management handled it.

Ganesh’s first-person account discusses ambiguity, interrupted focus, and boundaries as reasons the role can be a poor fit. Treat it as a useful perspective to investigate with prospective teammates, not a universal description. Vinoo Ganesh

A practical deployment playbook

Seven working habitsEditorial recommendations

Use these habits to turn technical work into something a customer can adopt and operate. Adapt them to the team, domain, and contract. They are recommendations, not a universal description of the role.

Evidence to carry from one stage to the next

Use these checks with the field workbook. They combine the two readings into a practical review; the gates are editorial exercises.

Stage	Bring to review	Pause when
Discover	Operator evidence, stakeholder map, baseline, and assumptions.	A critical claim or decision owner is missing.
Frame	Approved scope, acceptance examples, dependencies, and keep/change/remove decisions.	The outcome cannot fit the available time, budget, or authority.
Integrate	API contracts, current source versions, access tests, and safe replay.	Authority depends on a prompt or a required policy is unavailable.
Evaluate	Held-out cases, failure severity, grader review, cost, and resulting state.	A high average conceals an unacceptable failure.
Launch	Closed prototype gaps, pilot limits, monitoring owner, and tested rollback.	A release-critical behavior is still mocked or unverified.
Handoff	Runbook, system overview, current specifications, ownership drill, and support boundary.	The receiver cannot recover or change the system.

Reading: Buehling, Chapters 4–8, 10, 13–17; Agent Factory specification course; FDE AF Model.

An adaptable first 90 days

Period	Focus	Exit evidence
Days 1–30	Learn the domain and workflow; get access; establish a baseline; ship a small safe improvement.	A user-reviewed scope, known risks, and one working change.
Days 31–60	Deliver a bounded production pilot with evaluation and monitoring.	Actual usage, known failure modes, and an agreed go / no-go decision.
Days 61–90	Improve adoption, hand off operations, and generalise one useful pattern.	A named owner, a tested runbook, and a documented result.

This schedule is an editorial planning template. Shorter engagements may compress it; regulated or access-constrained work may take longer. Ganesh describes a more observation-heavy first month in his own onboarding advice. Choose a plan with your manager rather than treating one practitioner’s calendar as a rule. Vinoo Ganesh

Security belongs in the first design

Separate trusted instructions from customer documents and other untrusted content.
Enforce permissions in the tools and downstream systems; do not ask the model to be the access-control boundary.
Limit tools, operations, and credentials to the workflow’s actual needs.
Require review for high-impact actions and maintain an auditable record of changes.
Test cross-user access, injected instructions, and repeated side effects before launch.

OWASP identifies excessive functionality, permissions, and autonomy as sources of agent risk. It also explains that retrieval and fine-tuning do not eliminate prompt injection. OWASP Gen AI Security Project OWASP Gen AI Security Project

Entry routes and career choices

Routes into the roleChoose evidence to build next

Experience requirements differ sharply. Apply to the level and domain that match your evidence, and use the roadmap to close specific gaps. Completing a checklist does not substitute for experience or guarantee an offer.

Choose your starting point

Background	Useful existing strength	Next proof to build
New graduate	Recent engineering foundations and time for a complete project.	Target explicitly junior roles; show a working integration, tests, and honest user feedback.
Backend / full-stack engineer	Production code, debugging, APIs, and system design.	Lead discovery with a user, agree the outcome, and demonstrate adoption and handoff.
Data engineer / scientist	Data modelling, pipelines, analysis, and evaluation.	Own an application interface, permission boundaries, serving, and operational recovery.
Solutions engineer / consultant	Customer communication, domain knowledge, and scoping.	Show code you personally maintained after go-live, with tests and production incidents.
Founder / early employee	Broad ownership and fast iteration.	Explain depth, trade-offs, reproducibility, and how you collaborate in a larger delivery team.

Early-career routes exist, but senior roles are not interchangeable

Palantir’s New York commercial new-graduate FDSE posting targets graduates in December 2026 or spring 2027. It is a concrete early-career route with its own eligibility window. Palantir

The current US Databricks AI FDE advert explicitly excludes internship, new-graduate, and entry-level applicants. Do not read “all levels” in a heading as permission to ignore those requirements. Databricks

Decide what you want the role to teach you

To move toward product engineering, retain evidence of design, tests, code review, maintainability, and a component used beyond one customer. To move toward applied AI, document evaluations, data decisions, and operational trade-offs. For deployment leadership, show how you scope, staff, prioritise, and hand off work. For founding, record what you learned about buying decisions and repeatable customer needs.

These are career-planning recommendations. A small profile sample cannot establish the probability of any exit, and visible alumni founders are a selected group rather than evidence that the role causes startup success.

Who hires forward deployed engineers

Fifty-seven employersGlobal, dated evidence

A global research shortlist for engineers pursuing FDE work. Twelve employer pages were re-read on 14 September; older rows are preserved as 10 September snapshots and require a fresh check. “Indexed lead” means the employer’s advert was visible in search results but its current open status could not be independently confirmed.

Global field map: follow the requisition, not the company

The FDE market is international but not location-agnostic. A company with a global careers page can still hire a particular role only in one city. Check the legal employer, eligible countries, work authorisation, language, office cadence, customer travel, clearance, time-zone coverage, and employment-versus-contract terms before investing in a long interview process.

Representative employer evidence retrieved 14 September 2026. This is a location map, not a vacancy census.
Region	Current primary-source examples	What engineers should check
North America	Anthropic lists its FDE role in New York, San Francisco, and Seattle; OpenAI and Palantir also publish city-specific US roles.	State or city pay bands, hybrid attendance, travel, and clearance or citizenship requirements on government work.
Europe	Palantir London and Anthropic Munich are direct FDE-family examples; current boards also show Paris, Madrid, Amsterdam, Stockholm, and other hubs.	Country eligibility, local language, travel across borders, hybrid cadence, and whether “Europe remote” names your country.
Asia-Pacific & India	OpenAI Tokyo, OpenAI Sydney, FourKites India, and Razorpay Bengaluru show several distinct local markets.	Required working languages, relocation, hybrid presence, domestic customer travel, seniority, and whether “remote India” is India-only.
Latin America	Telnyx names Bogotá, Mexico City, and São Paulo; Handoff publishes a São Paulo hybrid role; Caylent publishes a Mexico role.	Spanish or Portuguese requirements, regional travel, local work authorisation, contractor status, and USD-versus-local-currency terms.
Middle East	OpenAI Abu Dhabi uses the adjacent Applied AI Engineer title for end-to-end customer deployment; Palantir’s board lists forward-deployed roles in Abu Dhabi and Dubai.	Exact title and code ownership, relocation, fixed-term status, client-site expectations, sector requirements, and regional travel.
Africa	Sand Technologies publishes a Kenya-titled, permanently embedded FDE role in public-sector health delivery; one sentence inside the posting says Malawi, so confirm the country directly.	Exact deployment country, long-duration relocation, local language and domain knowledge, public-sector work, infrastructure constraints, and ownership after deployment.

Search globally without losing precision

Search for forward deployed software engineer, customer engineer, deployed engineer, agent deployment engineer, applied AI engineer, implementation engineer, and partner deployed engineer. Then assess code ownership and production responsibility. An adjacent title can be a good match; a title match can still describe a different job.

The directory’s older rows are research leads, not fresh hiring assertions. Their travel and pay cells also belong to the older snapshot. Compensation is left in the employer’s published currency and format because exchange rates do not make employment terms comparable.

Compensation: compare the role, location, and pay components

Seven employer rangesChecked 14 September 2026

A posted range is evidence about one requisition. It is not an offer, a median, or total compensation. These examples were read from employer pages or their job-board API; the table preserves the employer’s wording and identifies ambiguities.

Annual USD ranges observed 14 September 2026. No exchange-rate conversions or equity valuations are included.
Exact role and source	Published range	How to read it
OpenAI · FDE, SF	$185,000–$300,000	Posted compensation; equity offered separately.
OpenAI · FD Software Engineer, SF	$185,000–$325,000	Posted compensation; equity offered separately.
OpenAI Deployment Company · FDE, US	$170,000–$400,000	Posted compensation; equity offered separately. Different employing company.
Anthropic · FDE, NYC / SF / Seattle	$280,000–$320,000	Listed as annual salary; generic sales-role OTE boilerplate also appears. Confirm the cash/variable split.
Databricks · AI Engineer — FDE, US	$152,900–$210,155	Posted local range; boilerplate distinguishes salary and OTE by commission eligibility. Confirm classification.
Palantir · FDSE, New York	$135,000–$200,000	Salary range; stock units, sign-on, and other incentives described separately.
Palantir · New Grad FDSE, New York	$135,000–$145,000	Salary range; stock units and other incentives separate. Specific graduation cohort.

Five kinds of salary evidence

Source	Useful for	Does not establish
Employer advert	The budgeted range, location, and explicit pay components of that role.	What you will be offered, or how other roles are paid.
Government labour-condition filing	A dated employer attestation for a particular job and work location.	An observed payslip, all-in compensation, or a representative market median.
Employee or offer submission	A concrete reported package when level, date, location, and components are recorded.	A reliable median without enough independent comparable observations.
Aggregator estimate	A broad initial lead to investigate.	A confirmed FDE-specific package or a consistent job definition.
Modelled report	Scenario analysis if assumptions and input data are disclosed.	A survey of real people simply because the headline gives a sample size.

H-1B labour-condition data contains wage attestations; it is not payroll. The Department of Labor explains the applicable actual-versus-prevailing wage obligation. This guide’s earlier $155,000 median over 196 records is withdrawn as a decision statistic because its exact extraction, date window, and deduplication are not available here. US Department of Labor

Offer questions worth taking to the recruiter

Which legal entity, location, and level is this offer for?
What is guaranteed annual cash, what is target variable pay, and what determines payout?
What equity instrument is offered, on what vesting schedule, and with what liquidity and exercise conditions?
How are travel days, weekend travel, on-call coverage, expenses, and time off handled?
Is the published range salary or OTE? Which components are outside it?
For a contractor role, what services, time off, equipment, and insurance does the contract include?

A larger federal-sector band alone does not isolate a “clearance premium”: the assignments, location, scope, and level can differ. Likewise, comparing unrelated field and product adverts cannot establish that one career track pays less.

How automation changes the work

Evidence and interpretationNo forecast of job replacement

Coding tools can reduce implementation effort while increasing the importance of evaluation, integration, and ownership. Evidence that a task can be automated does not establish how many jobs will disappear. Use the comparison to examine your assumptions.

Tasks and delivery methods are changing

Choose the observations relevant to your work.

Human accountability remains in the advertised role

These observations do not prove permanent immunity.

Your selected evidence

A practical response

Learn to inspect generated code, test outcomes, design narrow permissions, and diagnose production failures. Keep a record of where AI accelerates the work and where it creates rework. Use that evidence to decide what to automate next; do not assume that tool adoption is either proof of job security or proof of replacement.

Your resource library

Find the right material for the work in front of you. Books, courses, engineering guides, and practitioner conversations selected for learning and delivering as an FDE.

Curated resourcesFree & paid optionsChecked 14 September 2026Read the research & recommendations

Editorial order: useful starting points first, then depth. Open a resource’s notes for prerequisites, access, and a practice task.

“Free content” includes resources with a substantial free route and optional paid editions. Course access does not cover cloud, model, or lab usage unless stated. Check current checkout terms before paying. Saved items stay in this browser.

Sources, evidence limits, and corrections

Source register109 sources · 108 web references and 1 supplied book

Current employer observations, official announcements, implementation guidance, and editorial recommendations serve different purposes. Each reference below identifies its publisher and access date. Publication dates and access limits are included where available. Research cut-off: 14 September 2026.

How to use the evidence

Employer pages are primary evidence of advertised requirements and ranges; they are not independent evidence of working conditions or realised pay. Company announcements establish what was announced, not whether every planned investment or acquisition has completed. Practitioner and investor essays provide a point of view. This is a curated guide, not an exhaustive census.

New research is linked at the point of use. Older directory rows retain their 10 September snapshot label and are not certified as currently open. Search-index-only leads are identified separately. No result from a search is treated as proof that a company or region has stopped hiring.

Read the research memo · Download the source register

Panaversity. The AI Agent Factory: About.Living web resource · Read 14 September 2026; used for curriculum and exercises. Ecosystem, earnings, and labor-market claims not independently validated.
Panaversity. Spec-Driven Development.Living web resource · Read 14 September 2026; used for curriculum and exercises. Ecosystem, earnings, and labor-market claims not independently validated.
Panaversity. The FDE AF Model.Living web resource · Read 14 September 2026; used for curriculum and exercises. Ecosystem, earnings, and labor-market claims not independently validated.
Panaversity. Designing the Vertical System of Record.Living web resource · Read 14 September 2026; used for curriculum and exercises. Ecosystem, earnings, and labor-market claims not independently validated.
Panaversity. Eval-Driven Development for AI Employees.Living web resource · Read 14 September 2026; used for curriculum and exercises. Ecosystem, earnings, and labor-market claims not independently validated.
Richard Buehling / Benmore. Forward Deployed: Discovery, AI-Driven Development, and the Forward Deployed Engineer.2026-02 · Inspected locally 14 September 2026. Chapters 4–8, 10 and 13–17 inform discovery, specification, release, handoff and support exercises. Original PDF not hosted; no public download verified.
Printed chapter guide: discovery pp. 18–56; development pp. 63–71; handoff pp. 77–81; support pp. 82–90. Read the book notes and evidence limits.
OpenAI. OpenAI launches the OpenAI Deployment Company.2026-05-11 · Primary source. Read 14 September 2026.
OpenAI. Introducing Frontier Alliances.2026-02-23 · Primary source. Read 14 September 2026.
Anthropic. Building a new enterprise AI services company.2026-05-04 · Primary source. Read 14 September 2026.
AWS. AWS GovCloud (US) Newsletter, July 1–15, 2026.2026-07 · Primary source. Read 14 September 2026.
OpenAI. Forward Deployed Engineer (FDE) — SF.Publication date not stated · Primary source. Read 14 September 2026.
OpenAI. Forward Deployed Software Engineer — SF.Publication date not stated · Primary source. Read 14 September 2026.
OpenAI. Forward Deployed Engineer (FDE), Healthcare — SF.Publication date not stated · Primary source. Read 14 September 2026.
OpenAI Deployment Company. Forward Deployed Engineer — US, New York.Publication date not stated · Primary source. Read 14 September 2026.
Anthropic. Forward Deployed Engineer — NYC / SF / Seattle.Publication date not stated · Primary source. Read 14 September 2026.
Anthropic. Forward Deployed Engineer — Munich.Publication date not stated · Employer posting. Read via employer Greenhouse API, 14 September 2026.
Databricks. AI Engineer — FDE, United States.Publication date not stated · Employer posting. Read via employer Greenhouse API, 14 September 2026.
Palantir. Forward Deployed Software Engineer — New York.Publication date not stated · Primary source. Read 14 September 2026.
Palantir. FDSE, New Grad — Commercial, New York.Publication date not stated · Primary source. Read 14 September 2026.
Palantir. Forward Deployed Software Engineer — London.Publication date not stated · Primary source. Read 14 September 2026.
OpenAI. Forward Deployed Engineer — Tokyo.Publication date not stated · Primary employer posting. Read 14 September 2026; Tokyo hybrid role, Japanese and English required, travel mainly within Japan.
OpenAI. Forward Deployed Engineer — Sydney.Publication date not stated · Primary employer posting. Read 14 September 2026; Sydney hybrid role, relocation assistance stated, travel up to 50%.
OpenAI. Applied AI Engineer — Abu Dhabi.Publication date not stated · Primary employer posting. Read 14 September 2026; adjacent title with end-to-end customer deployment responsibilities.
Caylent. Forward Deployed Engineer — Mexico.Publication date not stated · Primary employer posting. Read 14 September 2026; company describes itself as remote across Canada, the US, and Latin America.
Handoff. Forward Deployed Engineer — São Paulo.Publication date not stated · Primary employer posting. Read 14 September 2026; hybrid role, English interview, contractor status, and USD compensation stated.
Telnyx. Forward Deployed Engineer — LATAM.Publication date not stated · Primary employer posting. Read 14 September 2026; names Bogotá, Mexico City, and São Paulo, professional Spanish or Portuguese plus English, and 10–30% regional travel.
FourKites. Senior Customer Engineer / Forward Deployed Engineer.Publication date not stated · Primary employer posting. Read 14 September 2026; Chennai or remote-India role with discovery, integration, agentic workflow, and production responsibilities.
DevRev. Forward Deployed Engineer — Implementation Partners.Publication date not stated · Primary employer posting. Read 14 September 2026; Bangalore role spanning software engineering, systems integration, applied AI, and partner delivery.
Razorpay. Forward Deployed Engineer — Bengaluru.Publication date not stated · Primary employer posting. Read 14 September 2026; production engineering for strategic merchants, with seniority and payments context stated.
Sand Technologies. Forward Deployed Engineer — Kenya.Publication date not stated · Primary employer posting. Read 14 September 2026; permanently embedded public-sector health role, 12–24+ months and four client-site days weekly. The Kenya-titled posting contains one sentence referring to Malawi; applicants should confirm the deployment country.
Sierra · Vijay Iyengar, Arya Asemanfar, Angie Wang. The AI-native interview.2026-04-22 · Primary source. Read 14 September 2026.
Anthropic. Guidance on Candidates’ AI Usage.2025-07-10 · Primary source. Read 14 September 2026.
Anthropic. Demystifying evals for AI agents.2026-01-09 · Primary source. Read 14 September 2026.
Anthropic. Effective context engineering for AI agents.2025-09-29 · Primary source. Read 14 September 2026.
Anthropic. Building effective agents.2024-12-19 · Primary source. Read 14 September 2026.
MCP · David Soria Parra and Den Delimarsky. The 2026-07-28 Specification.2026-07-28 · Primary source. Read 14 September 2026.
Model Context Protocol. C# SDK Versioning.Publication date not stated · Primary source. Read 14 September 2026.
US Department of Labor. H-1B Workers.Publication date not stated · Primary source. Read 14 September 2026.
Henley Wing Chiu · Bloomberry. What I learned analyzing 1K forward deployed engineer jobs.2026-01-25 · Original analysis. Originally published 18 November 2025; one analyst, 1,000 postings and 100 profiles.
Marc Andrusko · Andreessen Horowitz. The Palantirization of everything.2026-01-16 · Investor analysis. Read 14 September 2026.
Palantir Developer Community · PCL team guide. Writing Effective Evals.2025 · Practitioner guidance. Read 14 September 2026.
Palantir. Ontology design: Validation.Publication date not stated · Primary source. Read 14 September 2026.
Coder. Senior Partner Deployed Engineer.Publication date not stated · Indexed employer posting. Search-index text retrieved 14 September 2026; direct page requires JavaScript. Opening status not independently confirmed..
TechTorch. Forward Deployed AI Engineer.Publication date not stated · Indexed employer posting. Search-index text retrieved 14 September 2026; direct page requires JavaScript. Opening status not independently confirmed..
BLP Digital. Forward Deployed Engineer.Publication date not stated · Indexed employer posting. Search-index text retrieved 14 September 2026; direct page requires JavaScript. Opening status not independently confirmed..
ElevenLabs. Deployment Strategist Lead — Poland.Publication date not stated · Indexed employer posting. Search-index text retrieved 14 September 2026; direct page requires JavaScript. Opening status not independently confirmed..
OpenAI. Interview guide.Publication date not stated · Employer guidance. Read 14 September 2026; general guide, not a guaranteed FDE loop.
OWASP Gen AI Security Project. LLM06:2025 Excessive Agency.2025 · Practitioner guidance. Read 14 September 2026.
OWASP Gen AI Security Project. LLM01:2025 Prompt Injection.2025 · Practitioner guidance. Read 14 September 2026.
Vinoo Ganesh. The Definitive Guide to Forward Deployed Engineering.2026-02-05 · Practitioner guidance. Read 14 September 2026; Ganesh is a first-person account, not population evidence.
Harvard. CS50’s Introduction to Programming with Python.Learning resource checked 14 September 2026. Curriculum selection is editorial.
Scott Chacon & Ben Straub. Pro Git.Learning resource checked 14 September 2026. Curriculum selection is editorial.
MDN. Learn web development.Learning resource checked 14 September 2026. Curriculum selection is editorial.
University of Helsinki. Full Stack Open.Learning resource checked 14 September 2026. Curriculum selection is editorial.
PostgreSQL. PostgreSQL tutorial.Learning resource checked 14 September 2026. Curriculum selection is editorial.
FastAPI. FastAPI tutorial.Learning resource checked 14 September 2026. Curriculum selection is editorial.
FastAPI. Testing FastAPI applications.Learning resource checked 14 September 2026. Curriculum selection is editorial.
Government Digital Service. Using in-depth interviews.Learning resource checked 14 September 2026. Curriculum selection is editorial.
Hugging Face. AI Agents Course.Learning resource checked 14 September 2026. Curriculum selection is editorial.
Docker. Get started with Docker.Learning resource checked 14 September 2026. Curriculum selection is editorial.
OpenTelemetry. Observability signals.Learning resource checked 14 September 2026. Curriculum selection is editorial.
Google. The Site Reliability Workbook.Learning resource checked 14 September 2026. Curriculum selection is editorial.
MIT CSAIL. The Missing Semester of Your CS Education (2026).2026 · Checked 14 September 2026. Read the official course home and 2026 lecture index. The current course is the 2026 offering; older annual archives remain available.
Hamel's Blog. AI Evals: Everything You Need to Know.2025-05-28; updated 2026-09-01 · Checked 14 September 2026. Full public FAQ read, including publication date, 1 Sep 2026 modification date, error analysis, judge validation and production sections.
Latent Space. How Cursor deploys AI inside the enterprise.2026-07-01 · Checked 14 September 2026. Original interview read on publisher site; guest role, author and publication date verified. Company claims kept attributed rather than generalized.
Chip Huyen's Blog. Agents.2025-01-07 · Checked 14 September 2026. Original author article read, including date, table of contents, book-adaptation statement and experimental-framework caveat.
O’Reilly. AI Engineering.2024-12 (digital edition) · Checked 14 September 2026. Read the publisher overview, contents, and metadata. O’Reilly lists December 2024 for this digital title; this is the date recorded here. Paid full text was not accessed.
O’Reilly. Designing Machine Learning Systems.2022-05 · Checked 14 September 2026. Read O’Reilly's official title metadata and overview. The May 2022 scope is complementary to AI Engineering, not a replacement for its foundation-model focus. Description is based on public publisher/author material, not a full-text review.
O’Reilly. Hands-On Large Language Models.2024-09 · Checked 14 September 2026. Read the publisher metadata and the official https://github.com/HandsOnLLM/Hands-On-Large-Language-Models README/contents. Did not access paid full text or execute notebooks.
Hugging Face. Hugging Face Agents Course.Publication date not stated · Checked 14 September 2026. Read the official introduction, syllabus, prerequisites and certification terms. It describes a living course and requires a free Hugging Face account for Hub/Spaces activities.
The Full Stack. LLM Bootcamp.2023 · Checked 14 September 2026. Read the official syllabus, lecture descriptions and explicit free-access statement; did not watch every lecture or run the historical demo.
Model Context Protocol. Model Context Protocol specification — 2026-07-28.2026-07-28 · Checked 14 September 2026. Read the versioned specification overview, base protocol, features, extensions and security principles; checked official July release information. This entry does not claim every SDK implements every extension.
Lenny's Podcast. Why AI evals are the hottest new skill for product builders.2025-09-25 · Checked 14 September 2026. Original episode page, public description and timestamped chapters read; official outbound YouTube link verified. Paid transcript was not accessed.
LinkedIn Learning. Becoming a Forward Deployed Engineer: Bridging AI Products and Customer Outcomes.2026-08-10 · Checked 14 September 2026. Official course landing page and full public syllabus read; instructor, intermediate level, 51-minute duration, release date and locked lessons verified.
Will Larson / StaffEng. Staff Engineer: Leadership beyond the management track.2021 · Checked 14 September 2026. Official book page read, including free-content FAQ, guide links, interview basis and contextual limitation; prices intentionally omitted.
The Pragmatic Engineer. The Pragmatic Engineer.Publication date not stated · Checked 14 September 2026. Current official about page read: author, editorial scope, free versus paid access, schedule and subscription conditions verified.
Microsoft Learn. Anti-Corruption Layer pattern.Publication date not stated · Checked 14 September 2026. Primary engineering article read; canonical AWS PDF used where the newer landing page did not expose the article text.
Carnegie Mellon University. CMU 15-445/645: Intro to Database Systems (Spring 2026).2026 · Checked 14 September 2026. Read Spring 2026 home, complete schedule (updated April 27), and syllabus. Chosen over the in-progress Fall 2026 offering; video destinations are linked but playback was not verified.
DataTalks.Club. Data Engineering Zoomcamp.Publication date not stated · Checked 14 September 2026. Read the maintainers' current repository README, prerequisites, access table, and syllabus. It advertises January 2027 as the next cohort; live cohort means coordinated deadlines, not live lectures.
Microsoft Learn. Design and develop a RAG solution.Publication date not stated · Checked 14 September 2026. Read the primary architecture introduction, design phases, evaluation recommendations and update date of 30 Jun 2026. Did not read every linked phase article.
O’Reilly. Designing Data-Intensive Applications, 2nd Edition.2026-02 · Checked 14 September 2026. Read the publisher's second-edition page and contents: February 2026, Kleppmann and Riccomini. The older dataintensive.net marketing page still describes the earlier author/version. Description is based on public publisher/author material, not a full-text review.
O’Reilly. Fundamentals of Data Engineering.2022-06 · Checked 14 September 2026. Read the publisher overview and title page, which list June 2022 and the two authors. No claim that this is a newly updated edition. Description is based on public publisher/author material, not a full-text review.
Amazon Builders’ Library. Making retries safe with idempotent APIs.Publication date not stated · Checked 14 September 2026. Primary engineering article read; canonical AWS PDF used where the newer landing page did not expose the article text.
Martin Fowler. Strangler Fig.2024-08-22 (updated article) · Checked 14 September 2026. Primary engineering article read; canonical AWS PDF used where the newer landing page did not expose the article text.
Product Talk. Continuous Discovery Habits.2021-05-19 · Checked 14 September 2026. Author's dated book launch page and current Product Talk course/book navigation read; book scope and purchase access verified. Paid book text was not reviewed.
Latent Space. Every Agent Needs a Box — Aaron Levie, Box.2026-03-05 · Checked 14 September 2026. Official episode page, chapter list and transcript sections on identity, permissions and enterprise data read; publication date and public access verified.
Product Talk. Everyone Can Do Continuous Discovery—Even You! Here's How.2023-08-02 · Checked 14 September 2026. Original author article and edited conference transcript read; publication date, topic and free video/text access verified.
Google for Developers. Technical Writing Courses for Engineers.Publication date not stated · Checked 14 September 2026. Read Google's course home, audience requirements, and delivery-format guidance. Self-study lessons remain valuable without a facilitated session; no scheduled class or certification is promised.
Rob Fitzpatrick. The Mom Test.Publication date not stated · Checked 14 September 2026. Official author/book page read: scope, author, formats and purchase links verified; no publication date displayed and no price quoted. Practice is an editorial exercise. Description is based on public publisher/author material, not a full-text review.
Rob Fitzpatrick and Devin Hunt. The Workshop Survival Guide.Publication date not stated · Checked 14 September 2026. Official book page, authors section, complete table of contents and purchase link read; publication date not displayed. Full book not reviewed.
Simon Willison's Weblog. Agentic Engineering Patterns.2026-02-23 · Checked 14 September 2026. Official guide index, introductory chapter and dated launch post read; current index includes testing, review and code-understanding patterns.
O’Reilly / Cosmic Python. Architecture Patterns with Python.2020-03 · Checked 14 September 2026. Read the authors' Cosmic Python home and preface plus O’Reilly metadata. Authors explicitly offer free authorized reading; publisher confirms March 2020.
University of Helsinki. Full Stack Open.Publication date not stated · Checked 14 September 2026. Read the official home and https://fullstackopen.com/en/part0/general_info/. The course now updates continuously rather than using yearly editions; prerequisites and free access are explicit.
Google for Developers. Machine Learning Crash Course.Publication date not stated · Checked 14 September 2026. Read Google's course modules and prerequisites page; prerequisite page updated 25 Aug 2025. The course is a concept foundation, not deep API training.
O’Reilly / Google Abseil. Software Engineering at Google.2020-03 · Checked 14 September 2026. Read Google's Abseil landing page and book preface. Google confirms March 2020 publication and authorized free HTML availability.
Manning. The Design of Web APIs, Second Edition.2025-06 · Checked 14 September 2026. Read Manning's second-edition page, June 2025 metadata, reader prerequisites, chapter description, and linked code/exercise resources. Chosen over an older API-pattern catalog for its requirements-to-contract coverage. Description is based on public publisher/author material, not a full-text review.
Maven / Parlance Labs. AI Evals For Engineers & PMs.Publication date not stated · Checked 14 September 2026. Read the official public course syllabus and FAQ, which describe office hours, exercises, projects and lifetime material access. Full teaching content not accessed; no outcome guarantee or fixed fee asserted.
Microsoft Learn. Circuit Breaker pattern.Publication date not stated · Checked 14 September 2026. Primary engineering article read; canonical AWS PDF used where the newer landing page did not expose the article text.
OpenAI API documentation. Evaluation best practices.Publication date not stated · Checked 14 September 2026. Read the current public guide body, including its explicit Evals deprecation notice, design process, architecture examples and grader caveats. Not a recommendation to adopt the retiring API.
The Pragmatic Engineer. Incident Review and Postmortem Best Practices.2021-10-19 · Checked 14 September 2026. Original free article read, including date, author, survey limitations, incident sequence and distinction between free content and paid templates.
Made With ML / Anyscale. Made With ML.2023 · Checked 14 September 2026. Read the official course overview and setup lesson, including local no-GPU instructions and the 2023 citation. Did not validate every dependency or cloud workflow.
Software Engineering Radio / IEEE. SE Radio 706: Observability Tool Migration Techniques.2026-02-04 · Checked 14 September 2026. Original IEEE episode page, show notes and transcript read; episode number, guest, host, date and OpenTelemetry/migration scope verified.
Amazon Builders’ Library. Timeouts, retries, and backoff with jitter.2019 · Checked 14 September 2026. Primary engineering article read; canonical AWS PDF used where the newer landing page did not expose the article text.
OpenAI / Booking.com. Booking.com and OpenAI personalize travel at scale.Publication date not displayed · Checked 14 September 2026. Used for the traveller-journey gap and Trip Planner data connection; engagement claims are not independently audited.
Morgan Stanley. Morgan Stanley Wealth Management Announces Latest Game-Changing Addition to Suite of GenAI Tools.2024-06-26 · Checked 14 September 2026. Customer-owned launch announcement used for Debrief’s consent, outputs, Salesforce write, and advisor decision boundary.
AWS / CBRE. CBRE and AWS perform natural language queries of structured data using Amazon Bedrock.2024-05-30 · Checked 14 September 2026. Technical prototype used to illustrate integration boundaries; not presented as a current reference stack.
OpenAI / Morgan Stanley. Morgan Stanley uses AI evals to shape the future of financial services.Publication date not displayed · Checked 14 September 2026. Used for expert grading, meeting evaluation datasets, and daily regression testing; full data and rubrics are not public.
Anthropic / Doctolib. Doctolib accelerates developer productivity with Claude Code.Publication date not displayed · Checked 14 September 2026. Used for the 30-engineer pilot, rollout, and shared setup repository; productivity results are customer claims.
OpenAI / BBVA. BBVA puts AI at the core of banking with OpenAI.2026-06-11 · Checked 14 September 2026. Used for internal champions, workshops, governance involvement, and feedback channels; conflicting usage cadence labels are omitted.

14 September 2026

Added a Benmore and Agent Factory reading track, nine milestones, a field workbook, and six delivery gates. Sources distinguish author methods from measured outcomes. The supplied PDF is cited without hosting the original.

Expanded the employer directory and regional guidance for engineers worldwide, with primary-source examples from India, Kenya, Mexico, Brazil, wider Latin America, Japan, Australia, and the UAE.
Replaced unlinked growth and capital comparisons with official deployment announcements. Corrected DeployCo’s launch date, partner count, and proposed Tomoro headcount.
Rebuilt pay examples from seven employer requisitions; removed unsupported market medians and comparisons that conflated role, location, and seniority.
Changed the posting checker to acknowledge overlap with pre-sales and professional services. Its score is explicitly an editorial heuristic.
Added entry routes, Europe and Poland leads, production deliverables, evaluation measures, offer questions, and direct learning references.
Reframed interview content as verified policies plus original practice exercises. Added Sierra’s system-design screen and distinguished its debugging pilot from established rounds.
Corrected claims that indexed retrieval is obsolete and that older MCP peers can be ignored. Updated machine-readable descriptions to match the page.

Earlier change log · 10 September 2026

Historical editing notes below are retained for transparency, not as a current source. The 14 September corrections above supersede conflicting DeployCo, compensation, and technical claims.

10 September 2026

A full re-check of every checkable claim against primary sources — company job-board APIs, Levels.fyi, US Department of Labor filings, company engineering blogs and the protocol specifications. Thirteen corrections, several of them to figures this page had presented as verified.

A third pass removed a claim about an open semantic standard that no source supported, flagged that Anthropic’s published band does not say what is inside it, and added a note about how much of this field’s quantitative spine rests on one analyst’s single scrape. The pattern across all three passes is the same: the errors were not in the reasoning, they were in accepting a confident secondary summary without going to the source.
Every claim on this page was then re-checked a second time, adversarially, and eleven more things were wrong — several of them introduced by the first pass. The machine-readable answers in the page head were still serving the pre-correction figures to answer engines hours after the visible page had been fixed; a site whose whole argument is checking was quoting the version of itself it had just repudiated. Six pay-chart rows carried bands that matched no live requisition. Three roadmap claims about compliance deadlines had no source behind them and have been replaced with advice to read the current text rather than a summary. Details below.
The company directory was rebuilt from scratch, from 28 rows to 46, each read off the employer’s own Lever, Greenhouse or Ashby API on the day. Sierra, Decagon, Cognition and Writer had all renamed the role. Harvey, Glean, PostHog, Legora, Chainguard, Samsara and OpenEvidence are not posting a deployed title at all — they are kept in the table, marked, rather than deleted.
The pay section was regrouped by evidence quality rather than by size. The old headline row — “FDE title, all companies, $185–631k, median $259k” — was mislabelled: that range belongs to Palantir’s page. The genuine all-companies figures are a $205k median with quartiles at $175k and $277k.
The $610k and $1.2M figures now carry their source and its method. They come from a report published by a conversational-forms vendor whose own methodology describes the “1,200 FDEs” of its title as a synthesis of five public sources rather than a survey.
Employer-attested US Department of Labor filings were added as a compensation tier — 196 records for the exact title, median $155,000. No other resource in this niche uses them.
Anthropic, Databricks, Scale AI and Sierra moved out of the “not published” column. All four publish bands; the previous cells were wrong rather than merely empty.
Anduril’s valuation was two years stale. Corrected from $3.8B at roughly $30.5B to $5B at $61B, raised May 2026.
The “100+ companies founded by Palantir alumni” citation pointed at a dead page, and was re-sourced rather than dropped. The URL previously cited 404s, but the analysis is live elsewhere on the same domain: a Concept Ventures post counting 111 companies that had raised $11.6B by 2024, naming forward deployed engineer as the most common founder background. It now appears with that provenance and with its own caveats — no publication date, and an author who describes the analysis as partly AI-assisted.
The Anduril founding story was tidied up too neatly here. Trae Stephens came to it from a venture seat at Founders Fund, three years after leaving Palantir — not straight out of the field. Brian Schimpf, also ex-Palantir and now chief executive, was missing entirely.
Four errors in one DeployCo sentence: the $10B valuation is post-money not pre-money; five investors are named, not nineteen; Tomoro is an Edinburgh consultancy, not a London one; and it brought 95 engineers, not about 150. The 17.5% return is annual, over five years.
A correction that was itself wrong, reversed the same day. An earlier version of this entry claimed Anthropic’s “~25% travel” was really its office-attendance policy misread. It is not. The posting carries both figures separately — an estimated 25% travel to customer sites, and a separate policy of being in an office at least 25% of the time — and two coincidental 25%s on one page produced the confusion. The original figure stands, now scoped: 25% is the US req, while the Paris, Munich and manager reqs say 25–50%. Left here rather than deleted, because a changelog that only records other people’s errors is not a changelog.
The “one in three postings are pre-sales” claim was labelled “widely reported”. It is not. It traces to a single classification of 1,000 postings. It is now attributed, and paired with the caveat that the same dataset shows the quota red flag appears in under a tenth of postings — so it is a precise test, not a sensitive one.
Three roadmap skills were teaching outdated material. The MCP node described a protocol revision superseded in July 2026, when the protocol went stateless; the evals node recommended a platform being retired in November 2026; the retrieval node still treated pre-computed embedding search as the default. FedRAMP 20x and the revised EU AI Act dates were added.
Delta is not a synonym for FDSE. Palantir’s own job-board data shows it is the whole forward deployed engineering org, now running three ladders — including a Forward Deployed AI Engineer track that exists only in New York and London.

Common questions

What is a forward deployed engineer?

An engineer who works closely with customers to discover a problem, build and deploy software in its real operating environment, and support a measurable outcome. Exact scope, seniority, commercial duties, and travel vary by employer.

How is FDE work different from professional services or solutions engineering?

These categories overlap. Assess production code ownership, customer accountability, success measures, maintenance responsibility, and product feedback. A pre-sales task, OTE reference, or professional-services department does not by itself settle the role.

What does a forward deployed engineer earn?

The guide compares seven dated employer requisitions in annual USD, with links and pay-composition notes. Posted ranges are not realised compensation or a market median; equity, variable pay, seniority, location, and the employing entity require separate comparison.

What does the interview process look like?

The guide separates published employer policies from original preparation exercises. OpenAI describes variable assessments and round-specific AI permissions. Anthropic requires unaided assessments unless expressly allowed. Sierra describes a system-design screen and a plan, build, review onsite.

What should an FDE portfolio demonstrate?

A scoped customer workflow, functioning integration, evaluation suite, permission boundaries, failure handling, cost and latency measurements, user feedback, and an operational handoff. The project brief and suggested metrics are editorial recommendations.

Where are FDE roles available globally?

Current employer pages show FDE or closely related deployment roles across North America, Europe, Asia-Pacific, Latin America, the Middle East, and Africa. Coverage is uneven and role-specific. Remote eligibility, employment terms, language, travel, clearance, hybrid attendance, and current opening status must be checked for every requisition.