What an FDE Actually Does

The role was invented at Palantir in 2006. Twenty years later it is the most-discussed engineering title at OpenAI and Anthropic. Here is what it actually is.

One sentence that survives the hype cycle

A Forward Deployed Engineer is a software engineer whose primary workspace is the customer's environment, not the company's product backlog. The job is to make a generic platform fit one customer's situation closely enough that they keep paying. Nabeel Qureshi, an ex-Palantir FDE writing on Lenny's Newsletter in 2024, put it sharper: 'FDEs didn't merely empathize, they literally do their same job with your product as an extended member of their team.' Akshay Krishnaswamy, Palantir's Chief Architect, said the same thing from the inside in an April 2025 TBPN interview: 'the entire product development process runs front-to-back from the field, not from a privileged engineering core in Palo Alto.' That is the role. Everything else, the title, the org chart, the comp band, is downstream of that single fact.

Where it came from

Palantir invented and codified the term. Shyam Sankar joined Palantir in 2006 as employee #13 and became the first FDE, embedding with intelligence agencies and US military units, including deployments to Iraq and Afghanistan, to tailor the product to operational realities the company in Palo Alto could not have anticipated. Sankar is now Palantir's CTO. He is the architect of the entire model. Until roughly 2016, Palantir had more FDEs than it had 'normal' software engineers; the balance shifted only when Foundry launched as a more productized platform and the central engineering org grew faster. The Pragmatic Engineer's reporting and Qureshi's first-person essay both document this trajectory.

Dev vs Delta

Palantir's own engineering blog runs a post titled 'Dev versus Delta' that explains the internal split better than anything external. Devs build one capability that ships to many customers (the platform). Deltas, which is Palantir's internal name for the FDE function, deliver many capabilities to one customer (the deployment). The distinction matters because the optimization function is different. A Dev's success is measured by how well their feature generalizes. A Delta's success is measured by whether this particular customer renews. Trying to do both jobs in one role produces engineers who feel pulled in two directions; separating them is what makes the FDE model coherent.

What it is not

An FDE is not a sales engineer. Sales engineers exist to close deals; their artifact is a closed deal. An FDE may demo, but the artifact is shipped code that runs in the customer's production. An FDE is not a management consultant either. A consultant leaves a recommendation; an FDE leaves a running system. And an FDE is not a customer success manager whose job is to help customers use a finished product correctly. The FDE's product is intentionally unfinished at delivery, completed at the customer site. The research caveat: the popular narrative that 'FDE has been rebranded from consultant to AI engineer' was specifically refuted in adversarial verification for this report. Don't repeat it. The role is its own thing.

Why the role surged again in 2024-2026

For most of the 2010s the FDE pattern was associated mainly with Palantir and viewed by SaaS companies as a quirk of selling to government and finance. Foundation models changed that abruptly. A model that can read a customer's emails, write code in their codebase, or summarize their tickets is generic in a way no SaaS product ever was, and generic in exactly the way that makes FDE work valuable. OpenAI now runs a dedicated 'OpenAI Deployment Company', a $4 billion joint venture with TPG, Advent, Bain Capital, and Brookfield, and acquired the FDE-style firm Tomoro (which brought roughly 150 FDEs). Anthropic announced a $1.5 billion standalone FDE consulting joint venture with Blackstone, Hellman & Friedman, and Goldman Sachs in May 2026, with a $300 million founding commitment. Ramp runs about 15 FDEs in pods led by Leo Mehr. Google Cloud's FDE org has expanded aggressively under Thomas Kurian. None of these are coincidences. The pattern Palantir built for intelligence agencies turned out to be exactly the shape of selling foundation models to enterprises.

Who actually thrives

Sankar's own framing for FDE hiring, as quoted in Apoorv Agrawal's 2024 essay on Palantir's culture, is the most useful filter: 'I don't want to assemble a polite roster of cross-functional professionals. I want the X-Men, a medley of mutants united for good.' What he is selecting against is the safe, layered specialist who needs a clean ticket to start work. What he is selecting for is people who are technically dense but range widely, comfortable in ambiguity, willing to be the most technical person in the room and the least technical person in the room within the same day. Engineers who need long deep-work blocks, clean problem statements, and ownership of one system over years are usually happier as Devs. Neither shape is better; they're optimized for different jobs. The FDE pattern trades depth in any single codebase for breadth of context, and trades long-horizon ownership for fast feedback from real users.

Key takeaways

•FDE = engineer whose workspace is the customer's environment; the output is shipped software, not slides or recommendations.
•Invented at Palantir in 2006 by Shyam Sankar (employee #13, now CTO). Until ~2016, Palantir had more FDEs than traditional SWEs.
•Palantir's internal split: Devs ship one capability to many customers; Deltas (FDEs) ship many capabilities to one customer. Different optimization functions.
•Distinct from sales engineers (artifact = closed deal), consultants (artifact = recommendation), and CSMs (artifact = correct usage of a finished product).
•Surged in 2024-2026 because foundation models are generic in the way that makes customer-specific wrapping valuable. OpenAI, Anthropic, Ramp, and Google Cloud are all aggressive hirers.
•Sankar's hiring filter: 'X-Men, a medley of mutants', not 'a polite roster of cross-functional professionals.' Pick the role based on that, not on the salary band.

Exercise

Pick a SaaS product you've used heavily (Salesforce, Notion, a vertical tool like Toast or Epic). Write down three workflows that company almost certainly cannot ship as product features because every customer needs them slightly differently. Those gaps are where FDE work lives. Then ask: would you rather build the generic version of one of those features (Dev), or build all three for a single customer (Delta)? The honest answer tells you which side of the org you belong on.

Self-check

1.Quote Qureshi's one-line definition of what FDEs do that 'mere empathy' can't capture. Why is that the load-bearing distinction?
2.Explain Palantir's Dev vs Delta split. Why does measuring an FDE the same way you measure a platform engineer break the model?
3.Name three companies running aggressive FDE motions in 2025-2026 and the structural form each one is using (JV, in-house pod, acquisition).
4.What is Sankar's 'X-Men' framing selecting for, and what is it selecting against?

Sources

Next lesson

Problem Framing in the Customer's Environment

Translating what the customer asked for into what they actually need, without becoming the person who says 'actually you don't need that.'

→