"Forward Deployed Engineer" means different things at different companies. One thing stays constant: you own a customer outcome and write the code to deliver it. But who they hire, how senior, and what they emphasize varies enough that you should read the specific job description before you prep. This page maps the differences from primary sources, and is honest about where public data thins out.
Each employer below also has a dedicated interview guide that maps what is publicly known about its loop to the drills to run: Palantir, OpenAI, Anthropic, Google Cloud, and startups. Open the one for the company you are targeting.
The role, and the roles next to it
Before the per-company differences, fix the distinction that trips people up. The Pragmatic Engineer draws it cleanly. A solutions architect is advisory: they rarely write code on a customer's infrastructure. A forward deployed engineer is hands-on: they write code directly on customer infrastructure and work with more ambiguity than a traditional cloud solutions architect. If a posting is really a sales-engineering or advisory role in an FDE costume, the tell is that it never asks you to ship.
Palantir — the original
Palantir invented the forward-deployed model (the FDSE / FDE title), and it remains the most-documented loop (the whole Interview Method leans on it). Two things distinguish it:
- It hires the broadest seniority band. Unlike the frontier labs, Palantir's forward-deployed org hires across experience levels, including engineers with, per The Pragmatic Engineer, as little as a year of post-college work. If you're early-career, this is the most open door.
- The loop centers on decomposition, not algorithms. Its signature round is the open-ended problem breakdown, where algorithm grinding alone falls short.
Comp (Levels.fyi, self-reported, current as of July 2026): the Forward Deployed Software Engineer title shows a median total compensation around $201K, with a range of roughly $159K–$325K in the US. Self-reported, so directional.
OpenAI — the deepest seniority ask
OpenAI's FDE posting asks for the most experience of the majors and frames the role around end-to-end ownership.
- Seniority: "5+ years" of engineering or technical deployment experience that includes customer-facing work.
- What they emphasize: owning the whole arc. The role is described as "owning discovery, technical scoping, system design, build, and production rollout," and contributing "directly in the code when progress depends on it." They want someone who has "scoped and delivered complex systems in fast-moving or ambiguous environments."
- Loop tell: a first-hand account describes an API take-home followed by a walkthrough (one first-person account). Corroborate against the live posting, since independent detail on the frontier-lab loops is thin.
Anthropic — MCP-native, customer-embedded
Anthropic's Applied AI / FDE role skews slightly less senior than OpenAI's and is distinctly shaped by its own protocol stack.
- Seniority: "3+ years" in a technical, customer-facing role, or as a software engineer with consulting experience.
- What they emphasize: embedding "directly with our most strategic customers" and shipping production artifacts in their ecosystem. The posting names delivering "MCP servers, sub-agents, and agent skills" used in production, and asks for "high agency with an ability to navigate ambiguity." If you're targeting Anthropic, knowing the Model Context Protocol is close to table stakes.
Google Cloud — the tiered ladder
Google moved into FDE hiring in 2026, and The Pragmatic Engineer decodes its posting as roughly a quarter coding, half integration and plumbing, and a quarter meetings and customer hand-holding, with "founder's mindset" translating to "no one will provide a spec, and scope creep is your problem." The same title spans multiple levels and bands, so go by the specific posting's years-of-experience line. The lived-mechanics picture is on The Job, Week to Week.
Startups — the versatility bar
At a startup, "forward deployed engineer" usually means "the engineer we send to the customer, who also does everything else." There's no separate ladder and no documented loop; the bar is versatility and ownership under less structure. Public, citable specifics are essentially absent, so judge any startup FDE role by what its job description actually asks for; the title tells you little.
What this means for your prep
- Early-career: Palantir's broad band is the most accessible entry, and its decomposition-first loop is the best-documented to prep against.
- Mid-to-senior with production scars: the frontier labs (OpenAI 5+, Anthropic 3+) reward demonstrated end-to-end ownership and, increasingly, fluency in their own stacks (MCP for Anthropic).
- Everyone: the evidence bar is the real differentiator across all of them. The seniority line gets you read. Shipped production work is what gets you hired.
Next: Real Deployments — Overview: the cycle you've been drilling, run in real companies.NextOverview