What does a Forward Deployed Engineer actually do?

The role was crystallised inside Palantir in the mid-2000s. Palantir's bet was that selling shrink-wrapped software to intelligence agencies, banks and hospitals would never work — every customer's data, ontology and workflow was too specific. So Palantir invented a role whose mandate was to land inside the customer's environment, learn their domain, and bend the product until it produced real outcomes. That role became the FDE.

Fifteen years later, OpenAI, Anthropic, Scale, Cohere and dozens of AI-native scale-ups have re-discovered the same pattern. LLM-powered products are inherently fuzzy: prompts, evals, retrieval, tool-use and guardrails must be tuned to each customer's data and risk posture. Shipping a generic SaaS dashboard doesn't cut it. The FDE is back — and this time the deployment cycle is measured in weeks, not years.

A senior FDE spends roughly a third of their week with the customer — on calls, in their Slack, occasionally on-site in New York or San Francisco. They map the customer's workflow, instrument the painful steps, and identify where the product can replace a human bottleneck. They write code that lives in two repos: the customer's, and the product's.

Another third is spent building. This is real engineering: TypeScript, Python, infrastructure-as-code, evals, retrieval pipelines, prompt scaffolds, custom UI surfaces. Unlike a Solutions Engineer, an FDE ships production code — sometimes into the customer's stack, sometimes back into the core product as a new feature.

The final third is feedback. The FDE is the highest-bandwidth channel between the customer and the product team. They write tight memos, file specific issues, and lobby PMs to fix what they saw break. The best FDEs are treated as honorary members of the product org.

Solutions Engineers are pre-sales: they demo, scope and help win the deal. Customer Engineers are post-sales: they onboard, train and renew. The FDE sits between and beyond both — they own a deployed outcome, not a pipeline stage. We unpack the comparison in detail in our companion piece on FDE vs Solutions Engineer.

A useful test: if the engineer is measured on logo-acquisition or ARR, they're an SE. If they're measured on usage, retention or a customer-specific KPI (mean time to triage, agent containment rate, throughput per analyst), they're an FDE.

LLM products fail in customer-specific ways: their data isn't clean, their workflows aren't documented, their compliance posture rules out half of the obvious solutions. Generic onboarding fails. A great FDE compresses the time-to-value from quarters to weeks by writing the integration, the evals and the change-management plan in parallel.

This is why founders building Forward Deployed Engineer teams in the US are paying $1,400–2,200 per day for proven operators. The ROI is brutal and obvious: a deployed account is worth 10–50x the FDE's annual cost.