A10XFDE training

AI Forward Deployed Engineer · In a loop now, or getting there

Become an AI Forward Deployed Engineer.

The labs shipping AI agents to enterprises pay senior money for engineers who can take one from demo to production. A10X is the training: a guided 14-day sprint you run at your own pace, an AI coach on every step, and timed interview cases graded on the rubric real loops use. You leave with a deployed agent, an eval suite, and a shadow-rollout writeup, the artifacts those loops actually screen for. Already interviewing? Start with the customer case round: write your answer, and the coach grades it before the real one does.

Join A10X · $249 onceTry a free interview case →

One payment · No subscription · 30-day money-back

Companies hiring Forward Deployed Engineers

2 Oct 2026

Anthropic will train 10,000 Frontier Deployed Engineers by the end of 2027, by nomination only. Not nominated? Here is what its graded practical tests, and the open path →

§01The opportunity

The best job in software is also the hardest to break into.

A Forward Deployed Engineer embeds with a customer to design, build, and ship AI systems in production, not a demo over the wall. It pays like senior product engineering (see FDE salary ranges), and the interview loop doesn't want a certificate. It wants proof you've shipped.

a course, 6 months laterCOURSE ENDSTODAY
01

FDE salary: senior money, and it's climbing

Every listed FDE role on our board pays a senior salary or better, and the top listed base salary runs to $400K. Equity, where a company offers it, comes on top, and pay varies by lab, level and location. Every frontier lab and AI startup is racing to staff the role.

02

They screen for artifacts, not credentials

The interview tests one thing: can you build an agent, prove it with evals, and roll it out safely? A cert can't answer that. A portfolio can.

03

Generic courses are slow and stale

A 32-week cohort teaches the field as it was when the syllabus was written, and the field moves weekly. You need speed and currency, not a lecture archive.

§02The sprint

How the sprint gets you there.

A guided path, not a content dump. Each block ends in a real artifact: the three that FDE interview loops screen for. Seen the 30-day FDE roadmapeveryone's sharing? This is that plan, compressed to 14 days, with every artifact graded instead of guessed at.

Days 1–2

Foundations & target

Map the role, pick your build, stand up the stack: Python, an agent loop, one cloud.

OUTPUT: project spec
Days 3–6

Build the agent

Ship a production-style agent: tool use, RAG over real data, error handling, a real integration.

ARTIFACT ①: deployed agent
Days 7–9

Prove it with evals

Golden dataset, regression suite, drift detection, tracked failure modes. The skill the loop actually tests.

ARTIFACT ②: eval suite
Days 10–12

Ship a shadow rollout

Run in shadow against real traffic, diff the baseline, write the staged-rollout plan.

ARTIFACT ③: rollout writeup
Days 13–14

The interview loop

Drill the customer case study, the lowest pass rate and highest weight round, until it's reflex.

OUTPUT: loop-ready

Already an FDE? Start at the eval + rollout blocks. The coach paces the sprint to where you are.

What that looks like in practice: we ran a full eval on a post-cutoff decision model and published it, scores, confidence calibration and every confident mistake included. Read the benchmark →

§Try it before you buy

Can you pass a real FDE interview case?

The customer case study is the highest-weight, lowest-pass-rate round in FDE loops. Try one free: a messy customer with hidden traps, on a real clock, and get graded by the coach against the rubric top labs actually use. No signup to start.

Try the free case →

Live rubric grade

Discovery & constraint mapping · 2/5
Jumped to a solution before mapping PHI + clinical-refusal rules.

Eval & safety plan · 1/5
CSAT measured after launch, i.e. after harm is possible.

§03The system

Everything pointed at one outcome: interview-ready.

Not a video library. FDE training as reps: drills, graded feedback, and the live roles you're aiming at.

// BUILD

Guided drills → artifacts

You ship, you don't watch. Build the agent, write the evals, run the rollout, and walk out with a portfolio, not notes. Rewritten continuously as the field moves.

// FEEDBACK

A coach inside every drill

An AI coach built for FDE prep. Embedded in each sprint day, so you ask questions in context, and it grades your agent, evals, and rollout writeup against checklists built from public FDE loops. Then drill the case-study round under a timer in the Arena.

// KNOW YOUR TARGET

Company dossiers + the jobs

Sourced intel, not guesses. Dossiers on the companies hiring FDEs: stack biases, how their loop actually runs, and how to shape your artifacts for each and a weekly stack-watch brief so your edge compounds after you're hired.

§04The jobs

FDE jobs, hiring right now.

This is what you're training for: live FDE jobs at the labs and startups building the frontier. Three below; the full board is on /jobs. New to the role? Start with what an FDE actually is.

Forward Deployed Engineer (Training)
Baseten
San Francisco
$200K–$400K
Senior / Staff Forward Deployed Engineer
Giga
San Francisco
$180K–$400K
Senior / Staff Forward Deployed Engineer
Legora
New York City
$237K–$369K
See all 367 open roles →Rebuilt from each employer's posting API every 12 hours · see the full board
§05Stay current

And after you're hired, you don't go stale.

The field moves weekly. Members get the brief on what changed and exactly what to do about it, so your edge compounds instead of decaying.

FROM: A10X FIELD DESKRE: BRIEF №12 · THREE MOVESDATE: NOVEMBER 2026 · WK 45
comp

One FDE posting, three pay ranges. The top one is in another country

On 2026-10-28 Cohere posted a Forward Deployed Engineer, Inference role, located in the United States. Its pay headline reads $240K to $415K. Under it sit three ranges: California, New York and Washington $265K to $325K; the rest of the US $240K to $295K; Canada $340K to $415K. So the $415K is the Canada range, not a US one. It is marked as US dollars, but it sits above both US ranges, and Cohere's other US FDE posting, Agentic Platform, prices its Canada range in Canadian dollars (CA$175K to CA$385K). Our own board showed $415K as its top base for a few hours on 29 October before we caught it. The role itself is infrastructure work: serving models inside the customer's own infrastructure, prefill and decode performance, vLLM or TensorRT-LLM, Kubernetes across clusters, 8+ years.

Before you quote a range, open the pay section of the posting and find the line for your state or country. If you could work from more than one place, ask the recruiter which range the role is budgeted in. Then check it against the same title on the live board, which now reads the range for the role's own country.
hiring

Three FDE employers, re-read on 28 October: four days in the office, 38 countries, no pay published

We re-read the hiring boards of three companies that hire FDEs in different ways. Ode, which Anthropic launched in July with Blackstone and Hellman & Friedman: 14 open roles, and every one that states a work arrangement (13 of 14) says 4 days in person, Fridays remote, in San Francisco, New York or Raleigh-Durham. DeployCo, OpenAI's deployment company: 30 open roles; the US FDE is New York or San Francisco, 3 days a week in the office, up to 50% travel, $170K to $400K plus equity, and it asks for 4+ years of engineering. Wonderful: 126 open roles, 39 of them titled Forward Deployed Engineer across 38 markets, none in the US, and 0 of the 126 publish pay.

Pick the one that fits where you live and how you want to work, and read its dossier: what the job asks, how the interview runs, and what we could not confirm. All three are free at a10x.dev/loops. If you go for Wonderful, ask on the first call for the pay range in your country, because no posting gives one.

Every brief ends in a drill. Read the full brief → · Field notes: the FDE role, salary, and how to become one →

§06Why A10X

Why not a course, or just ChatGPT?

Information about the field is free and everywhere. A path to a specific outcome, with feedback and the jobs attached, is not.

A10XFDE.AcademyGSDC CertJust ChatGPT
GETS YOU TOInterview-ready, at your paceDone in 32 wksA certificateAn answer
FORMATGuided sprint + coach32-week cohortSelf-paced videos + MCQ examYou drive, alone
PORTFOLIO ARTIFACTSAgent · evals · rolloutCapstoneOne capstoneYou assemble it
GRADED VS REAL LOOPSYes · public-loop checklistsCohort review—No
LIVE JOB BOARDIncluded———
STAYS CURRENTUpdated as the field movesEvolving, per cohortCert valid 5 yrsAs good as your prompts
COST$249 onceNot published (sales call)$300 promo, $600 list$20/mo
§07One payment

Pay once. Keep it for as long as A10X exists.

$249once

No subscription. Nothing renews. Everything below, and everything we add after.

Join A10X · $249 once →Interview in the next few weeks? 30 days for $49 ↓
  • ✓ The full 14-day sprint: agent, evals, shadow rollout
  • ✓ The AI FDE coach, inline on every day of the sprint
  • ✓ Company loop notes: how the loop runs at OpenAI, Anthropic, Salesforce, Palantir & more
  • ✓ The Case-Study Arena: timed reps on the round that fails most candidates
  • ✓ The Monday Brief with stack watch (the job board is free to everyone)
  • ✓ 30-day money-back, no questions
Join A10X · $249 once →

Secure Stripe checkout · Full refund in 30 days if it's not for you. Email hello@a10x.dev

Interview in the next few weeks?

Take the 30-day pass. You get everything in the $249 plan for 30 days: the sprint, the coach, the notes on how each company interviews and the graded cases. One payment of $49. Nothing renews.

Get 30 days · $49 →

Want to keep it after your interviews? Buy the $249 plan any time.

NOTE · Members who joined on the $39/month founding plan before 28 August 2026 keep that plan unchanged. Nothing about it renews differently.

§08Not ready?

Get the free Monday Brief.

The weekly what-changed / what-to-do for FDEs. Free, forever. See the quality before you commit; start the sprint whenever you're ready.

No spam · Unsubscribe anytime · ~1 email/week

§09Straight answers

Questions worth asking.

What is a Forward Deployed Engineer?

A Forward Deployed Engineer (FDE) is a software engineer who embeds with a customer to design, build, and ship production AI systems end-to-end (agents, evals, and safe rollout) rather than handing off a demo. Frontier labs staff FDEs to take models from prototype into enterprise production. Full explainer →

How much does a Forward Deployed Engineer make?

Forward Deployed Engineer roles pay senior salaries and above. On our live board the top listed base salary currently runs to $400K. Equity, where a company offers it, comes on top, and pay varies by lab, level, and location. See the full salary breakdown and real comp on the live job board.

What do I actually get for $249?

If your interview is days away: six timed case studies on the round that fails most candidates, a graded review on every one, and company loop notes covering stack biases, interview-loop formats, and artifact advice for OpenAI, Anthropic, Salesforce, Sierra, Palantir, Ramp, Decagon, Cognition, and Glean.

If you have weeks: the full 14-day guided sprint. You build an agent, write its evals, and ship a shadow rollout, which is the portfolio those loops ask about.

Either way: an AI FDE coach that grades anything you paste, including the case answers you write before a real loop, and the Monday Brief with a stack watch on those companies. The job board is free to everyone, member or not. One payment, no subscription; 30-day money-back guarantee. Want a taste first? One of the six cases is free, and so is Brief №12.

How much time does it take, and what do I need to start?

It depends on when you interview. If a loop is days away, skip the sprint and go straight to the cases: each one runs on a 45 to 90 minute clock and is graded when you submit. Nothing in them needs the sprint first.

If you have weeks, the full sprint is about 35 hours of guided work across 14 sprint days. Day 1 is a 45-minute decision; the other days run 1.5 to 3 hours each. You set the pace: two focused weeks if you can clear them, several weeks around a full-time job if you can't, and the coach holds your place.

Prerequisite: you can already code (any language; the sprint uses Python-family tooling). We don't promise a $400K offer; we promise the artifacts and reps that get you into the loop.

Can I join Anthropic's Frontier Academy instead?

Anthropic's Claude Frontier Academy, announced on 2 October 2026, is by nomination only: an organization asks its Anthropic account team, and the first cohorts come from firms like Accenture, Deloitte and McKinsey. If you are not nominated, the open path is the same work: practice on deployment cases you have not seen, get graded on them, and ship one real deployment. A10X is not affiliated with Anthropic. What the Academy tests →

Isn't this just stuff I could get from ChatGPT?

ChatGPT gives you information. A10X gives you a graded path to a specific outcome: scored feedback on the three portfolio artifacts the loop tests, and the live jobs on the other side. Different product: one is a search box, one is a graded path into the loop with the jobs attached.

I'm already an FDE. Is this just for beginners?

No. Working FDEs use it the same way: skip foundations, start at the eval and rollout blocks, and use the coach and Monday Brief to stay sharp as the field moves.

Is it really one payment?

Yes. $249 once, billed by Stripe, and nothing renews. You keep access for as long as A10X exists, including everything we add later. If it is not for you, email hello@a10x.dev within 30 days and we refund it in full. Interviewing in the next few weeks? The 30-day pass is $49 once and does not renew either.

§10Who is behind this
Armando Gonzalez

Armando Gonzalez

Two years building AI systems in production, forward-deployed work included, and the hiring side of FDE loops. I write every brief, source every dossier from the company's own pages, and read every graded review the coach produces. If something on this site is wrong, it is my mistake and I will fix it.

Companies hire me to run this loop inside their own systems: embedded FDE consulting. Everyone else can write to me at hello@a10x.dev; a human answers.

Get on the frontier. Fast.

A guided path to the portfolio FDE loops actually grade, at whatever pace your week allows. After that the coach keeps working: paste anything you build and get it graded, including the case answers you write before a real loop. One payment.

Join A10X · $249 once

30-day money-back · no subscription