5 Key Responsibilities of a Forward Deployed Engineer in AI

A Forward Deployed Engineer (FDE) in AI is a software engineer who deploys and owns a company's AI product — LLM apps, RAG pipelines, agents — inside the customer's own environment, and is accountable for making it work in production. The role is defined by where the work happens: not in your own codebase, but embedded with the client, against their data, for their stakeholders. It is also an AI-native role, not a generalist one: in Dexity's analysis of 399 US FDE job descriptions, 80% require at least one hands-on AI/ML skill.

1. What does an FDE build inside the client's environment?

The core of the job: analyze a client's problem and build a bespoke AI solution against their systems and data — integrations, RAG pipelines, and agent workflows. Anthropic's FDE listing asks for engineers who "ship advanced AI applications" and deliver technical artifacts like MCP servers and sub-agents in enterprise environments (Anthropic). The Pragmatic Engineer frames the scope as "similar to a startup CTO: you own end-to-end execution of high-stakes projects" (The Pragmatic Engineer).

JD backing: of the 399 US FDE JDs Dexity analyzed, Python appears in 60% and API/system-integration depth in 19% — the toolkit for building inside someone else's stack.

2. How does an FDE handle deployment and production scaling?

FDEs turn prototypes into secure, compliant, production-grade systems — handling data migrations, cloud setup, and the "last mile" that makes a demo actually run for a real customer. Anthropic calls this "white-glove deployment support … in enterprise environments" (Anthropic). Palantir's own account describes the FDE as the engineer who builds "the last mile of the product to work in production" — still writing and debugging production code, not just advising (Palantir).

3. How does an FDE troubleshoot live client systems in real time?

Client deployments break in ways your own product never did — different data, different infrastructure, edge cases no one scoped. FDEs debug live production, resolve schema conflicts, and tune performance under operational pressure. As OpenAI's Head of Forward Deployed Engineering, Colin Jarvis, put it: "what the customer describes in scoping doesn't match the data/system reality on the ground" (The Pragmatic Engineer). Handling that gap in real time is the responsibility that most separates FDEs from internal engineers.

4. How does an FDE lead stakeholders and technical decisions?

An FDE is the primary technical point of contact for the account: leading client workshops, translating business needs into specs, and looping internal teams in with deployment insights. This is why the role carries heavy customer exposure — Anthropic lists up to 50% travel embedding on-site with customer engineering teams (Anthropic), and practitioners describe splitting the week between building and customer calls/demos (Palantir).

5. What does FDE knowledge transfer and enablement involve?

Deployment isn't done at go-live. FDEs train the client's team, document the workflows, and provide the ongoing support that turns a one-time delivery into sustained adoption — so the customer can run and extend the system without the FDE in the loop.

What skills do FDE responsibilities require? (399 JDs)

Source: Dexity analysis of 399 LinkedIn FDE job descriptions · US market · 2026.

What are the seniority levels and pay for FDE roles?

Dexity scan of 191 open Forward Deployed Engineer roles across 35 companies, July 2026.

Disclosed pay bands (33% of postings) center on $146K–$240K (range $110K–$369K). The field is led by Palantir, Databricks, and OpenAI, and every role is client-facing.

What is a Forward Deployed Engineer NOT?

• Not an AI infrastructure / platform engineer. They build the internal serving and orchestration stack; the FDE builds in the customer's environment and owns the client outcome.
• Not a sales engineer. SEs support the deal pre-sale; the FDE lives post-deal, accountable for the system actually working in production.
• Not a solutions architect. SAs design the blueprint and hand it off; FDEs own the full hands-on implementation end-to-end.
• Not a heads-down IC. Client interaction and delivery ownership are the core of the job, not a tax on it.

Build the skill these responsibilities reward

Every one of these responsibilities rewards the same thing: the ability to ship production-grade AI inside a real environment. That's exactly what Ship Production Code with AI is built for — a focused course led by a practitioner shipping production AI at Microsoft, covering the context-engineering framework, spec-driven workflow, and MCP integrations that make AI systems hold up in production. You leave with a deployable project, not just theory.