What is a Forward Deployed Engineer?

A Forward Deployed Engineer (FDE) is a software engineer who embeds directly inside a customer's environment to design, build, and ship working systems in production, rather than shipping code and handing it off. The role blends deep technical work with constant customer contact: an FDE maps a messy real-world problem, builds the end-to-end solution against the customer's actual data and infrastructure, and stays until it runs in production. In 2025-2026, the title has become the default way frontier AI labs get their models into enterprises.

Where the role comes from

The FDE role was coined at Palantir, modeled on the idea of a forward-deployed soldier stationed in the field, ready for rapid response. The founding insight was simple and still holds: enterprise data is messy, legacy systems are unavoidable, and a working system requires an engineer embedded inside the customer's environment, not a demo thrown over the wall. Palantir built its business on FDEs (internally FDSEs) who sat on-site, learned the customer's domain, and shipped software that actually got used.

AI turned that niche consulting-adjacent role into one of the hottest jobs in tech. By April 2026 there were roughly 5,330 forward-deployed engineering job postings, a 729% year-over-year increase, as OpenAI, Anthropic, Google, Sierra, Scale, and dozens of application-layer startups raced to fill it (The New Stack, MarkTechPost).

What a Forward Deployed Engineer actually does

An FDE owns a customer outcome from problem statement to production. On a given week that looks like:

• Discovery and scoping. Sit with the customer, decompose a vague ask ("we want an AI agent for support") into a concrete, evaluable spec.
• Building end-to-end. Write production code: LLM pipelines, RAG retrieval, agent loops, API and data integrations, and the glue that connects them to legacy systems and messy data.
• Evals and guardrails. Build the evaluation harness that proves the system works and keeps working: golden datasets, regression suites, drift detection, tracked failure modes. This is now the core of the job at frontier labs.
• Deployment and rollout. Take the system to production, often via a shadow or staged rollout, then tune it against real traffic and edge cases.
• Feedback to product. Carry what breaks in the field back to the core engineering and research teams.

The defining trait is that FDEs talk to customers constantly. They sit at the intersection of customer-facing problem solving and hands-on systems implementation.

Why AI made the role explode

The generative-AI era exposed a gap: a great model means very little if no one can deploy it inside an enterprise with messy data, legacy systems, and urgent timelines. Foundation models are largely undifferentiated at the API; the value is created in deployment. Getting an agent from an impressive demo to a reliable production system requires someone who can wire it into real data, write the evals that make it trustworthy, and iterate on-site under pressure. That someone is the FDE.

The economics reinforce it. Labs that win enterprise accounts on the strength of successful deployments can charge more and retain longer, so they staff FDEs aggressively and pay them like senior product engineers.

FDE vs. solutions engineer vs. ML engineer

These roles overlap but are not the same:

• Solutions Engineer (SE): Primarily pre-sales. An SE supports the sales cycle with demos, proofs of concept, and technical answers. They influence the deal but usually don't own the production build. FDEs are post-sale and own shipping the real system.
• ML Engineer (MLE): Builds and trains models, owns pipelines and infrastructure, and rarely talks to customers. The simplest distinction in the field: MLEs rarely speak to customers; FDEs speak to customers constantly. FDEs typically consume models rather than train them, focusing on orchestration, retrieval, evals, and integration.
• Forward Deployed Engineer: Owns the customer outcome end-to-end in production. Deeper and more customer-embedded than an SE, more customer-facing and integration-focused than an MLE.

A useful mental model: the SE gets the customer to yes, the FDE gets the customer to production, and the MLE builds the model both of them rely on.

The bottom line

An FDE is the person who makes AI real inside a company. As models commoditize and every enterprise tries to put agents into production, the engineer who can embed, integrate, evaluate, and ship becomes the scarce resource. That scarcity is why the role went from a Palantir specialty to the job frontier labs compete hardest to fill.