A Forward Deployed Engineer (FDE) is a software engineer who works inside a customer's environment to design, build, and operate solutions on top of a vendor's product. Rather than shipping features for an abstract user from behind a Jira board, an FDE sits next to the people who actually use the system, learns their workflow in painful detail, and writes the code that turns a generic platform into something that solves the specific problem in front of them.
Palantir built the role first, in the early 2010s, when its customers (intelligence agencies, defense, then later finance, healthcare, and manufacturing) couldn't fully articulate what they needed and certainly couldn't share their data with a vendor's product team. Palantir's answer was to send the engineers to the data. Internally those engineers were called "Deltas" - execution-focused builders paired with "Echo" team members who acted as embedded domain experts. Until around 2016, Palantir reportedly had more Deltas than traditional software engineers on staff. The title has since escaped Palantir and is being copied, fast, across AI labs and enterprise software.
What an FDE Actually Does
The job is deliberately hybrid. In any given week an FDE might:
- Discover the problem on-site. Sit with end users, watch them work, map their data, and figure out the workflow that the original slide deck never captured.
- Prototype quickly. Stand up a useful slice of the solution in days, often against messy, half-documented production data.
- Write production code. Build connectors, transformations, data pipelines, microservices, dashboards, agents - whatever the use case demands.
- Configure and extend the platform. Use the vendor's product as a foundation and fill the gaps with custom code, scripts, and integrations.
- Carry the feedback loop home. Surface what's missing in the product, file the bugs, push for the fixes, sometimes write the patch.
- Own the outcome. Stay with the deployment until it is used, trusted, and producing measurable value - not just installed and demoed.
The deliverable isn't a deck or a recommendation. It's a working system, in production, owned by the customer, that wouldn't exist without the FDE. Palantir's own writing on the role notes that FDEs spend roughly a quarter of their time onsite with customers; the rest is heads-down engineering work back at base.
FDE vs Solutions Engineer vs Consultant
The titles in this space tend to bleed into each other. The distinctions are worth getting right.
| Solutions Engineer | Forward Deployed Engineer | Consultant | |
|---|---|---|---|
| Primary goal | Win the deal | Ship a working system | Deliver an engagement |
| Code they write | Demos and PoCs | Production systems | Varies, often advisory |
| Time horizon | Pre-sale | Months to years post-sale | Fixed-scope project |
| Reports to | Sales | Product or delivery org | Services firm |
| Success metric | Bookings | Customer outcomes, product adoption | Billable hours, sign-off |
| Feedback into product | Indirect | Direct and continuous | Rarely structured |
A solutions engineer's job effectively ends at "yes." An FDE's job starts there. A consultant is paid for time and advice; an FDE is paid for the system they leave behind and the outcome it produces.
Why FDEs Are Suddenly Everywhere
Two things happened at roughly the same time.
First, enterprise platforms got harder to deploy. Data lakehouses, vector databases, agent frameworks, observability stacks - they're flexible and powerful, but none of them deliver value out of the box. Somebody has to model the domain, wire up the integrations, and tune the system to the customer's data and risk tolerance. Sales engineers don't have the runway. Pure delivery consultants don't have the product context. The FDE pattern fills exactly that gap.
Second, AI happened. RAG systems, agentic AI, and LLM-powered workflows are tightly coupled to a customer's data, vocabulary, and tolerance for error. A real AI deployment is co-engineered with the customer. It's not a SKU you swipe a credit card for. That's why OpenAI, Anthropic, Cohere, and a long tail of AI startups have all started building FDE teams. OpenAI went further in October 2025 and stood up an entire "OpenAI Deployment Company," picking up the agent consultancy Tomoro to seed it with experienced FDEs. Anthropic's applied AI org, where FDEs sit, has reportedly been growing roughly 5x year-on-year.
The numbers are striking. Postings for forward deployed engineering roles grew by more than 800% between January and September 2025, with Salesforce alone publicly committing to a team of 1,000 FDEs. An analysis of more than a thousand FDE postings put the median disclosed base at around $173,000, with about 70% of listings also mentioning equity. The role pays like senior engineering because it does senior engineering work, with the added load of being customer-facing.
What Makes Someone Good at This
The role attracts strong full-stack engineers who happen to like being in front of customers. The recurring traits:
- Engineering depth. Data modeling, distributed systems, APIs, SQL and analytics, and enough infrastructure muscle to actually deploy what they build.
- Domain absorption. The ability to pick up a customer's industry fast enough to be useful in their own language inside a few weeks.
- Product taste. Knowing when to extend the platform with glue code versus when to push the product team to absorb the change.
- Communication. Walking executives through tradeoffs, sitting next to an analyst inside a notebook, and writing internal updates that don't waste anyone's time.
- A bias toward shipping. A working v0 in two weeks beats a perfect design doc in two months, every time.
- Comfort with ambiguity. The requirements you arrive with are almost never the requirements you ship against.
The best FDEs are not lesser product engineers who drifted into customer-facing work. They're engineers who are genuinely good at turning vague problems into running systems.
Where FDEs Show Up
- Data platforms. Embedding with customers to model their data, build pipelines, and stand up production analytics on ClickHouse, Snowflake, or a lakehouse.
- Search and observability. Tuning Elasticsearch, OpenSearch, and Grafana deployments to real production traffic and real ingest patterns, not the ones in the docs.
- AI labs and AI startups. Building RAG systems, agents, and evaluation harnesses on top of frontier models for a specific enterprise workflow.
- Vertical SaaS. Financial services, healthcare, defense, manufacturing - anywhere the platform is generic and the use case absolutely isn't.
- Specialist consultancies. Firms whose engagement model is forward-deployed by default, rather than a packaged SaaS sale with a services tail bolted on.
Why Companies Hire FDEs Instead of Going It Alone
A customer could, in theory, just hire their own engineers and do this work in-house. Plenty try. The reasons they end up bringing in FDEs anyway:
- Speed. An FDE who has shipped the same shape of system five times moves at a pace a generalist hire simply can't match starting from zero.
- Product depth. FDEs know the platform's quirks, edge cases, and roadmap in a way that's hard to replicate from outside.
- Risk transfer. A working deployment in 12 weeks with an FDE on the hook looks very different on a steering-committee slide from an 18-month internal program with unknown unknowns.
- Pattern flow. Good FDE teams quietly carry patterns - and anti-patterns - between customers, and feed the worst of them back into the product.
The FDE model earns its keep when the platform is powerful but generic, the customer's problem is sharply specific, and the cost of a slow or failed rollout is high. That description happens to fit most serious data and AI initiatives in 2026 - which is why the title keeps spreading well beyond Palantir.