FDE Postings Grew 729%. The Real Pool Hides Under Three Other Titles.

Sourcegraph's August 4, 2026 "Who is Hiring?" post on Hacker News lists "Agent Engineer" as a distinct role, right next to SWE and Tech Lead. The same week, Indeed data showed Forward Deployed Engineer postings up 729% year over year while Live Data Technologies clocked the annual figure at 1,165%. If your Boolean still starts and ends with "forward deployed engineer", you are searching a shrinking slice of a much larger reality.

The 729% number is real, but it is the wrong story

FDE demand did explode, but the "there are no candidates" panic is largely a title-fragmentation problem, not a supply problem. Indeed FDE postings went from 643 in April 2025 to 5,330 in April 2026. Paraform, using Live Data Technologies, put the peak YoY at 1,165%. AlphaSense saw FDE mentions in company docs climb 17% in six months.

Meanwhile the candidate pool grew roughly 50%. That gap is what everyone is writing about. But "the pool" is being measured against a single exact-phrase title that stopped being universal sometime in mid-2025, when AI-native employers started shipping the same job description under new names.

Here is what fragmented, and where:

• Palantir lineage kept "Forward Deployed Engineer" (still the dominant title at Palantir, Gecko Robotics, Modal, Roboflow, Cresta, Amp Code).
• AI-native startups launched with "Agent Engineer" (Sierra, Vapi, Henry AI), "AI Application Engineer," or "Agentic Engineer."
• Consulting ate the middle. QuantumBlack (McKinsey) and BCG require 8+ years of software and infra experience for what is functionally the same role, just badged as "technical consultant."
• Hyperscalers shipped variants like "Software Engineer, Agent" inside larger SWE ladders.

The job description barely moved. Sierra's Agent Engineer JD reads like a Palantir FDE post: partner directly with customers, build and ship AI agents, engage with all levels of the customer's business.

What Refolk's index actually shows

Refolk's index holds 2,654 US profiles currently titled "Forward Deployed Engineer" and only 148 US profiles holding the fragmented titles "Agent Engineer," "AI Application Engineer," or "Agentic Engineer." That roughly 18-to-1 ratio is the number every sourcer is quoting. It is also misleading.

The 148 is small, but it is disproportionately concentrated inside the exact employers you most want to poach from. Sierra alone accounts for 9 of 25 sampled Agent Engineer profiles, or 36%. The rest cluster at Vapi, Henry AI (YC S24), Fusion Risk Management, and Google DeepMind.

The top employers holding the canonical "Forward Deployed Engineer" title in Refolk's index are Palantir, Gecko Robotics, Northslope Technologies, Modal, Amp Code, Roboflow, and Cresta. Geography skews NYC and SF.

729%

Why the fragmentation is employer-driven, not candidate-driven

Candidates take whatever title their offer letter says, so the fragmentation is a recruiting-differentiation move by AI-native employers, and it is deliberate. Sierra publicly frames Agent Engineer as sitting at "the intersection of software engineering, frontier AI deployment, and the most pressing challenges of our customers." That is the FDE job description with the serial numbers filed off.

Two forces are driving the rebrand:

1. Scarcity signaling in pay bands. Staff Agent Engineer TLM at Sierra hits a $310K median base and $410K at the top. FDE base salaries on Paraform run $150K to $217K with a $183K midpoint. AI-natives can price the fragmented title higher because it reads as new-frontier work rather than legacy consulting-adjacent. A "$310K Agent Engineer" recruits better than a "$310K FDE" from candidates who associate the Palantir term with slide decks and DoD SCIFs.
2. Distancing from the Palantir cultural brand. Sierra, Vapi, and Henry AI want to hire from OpenAI, Anthropic, and DeepMind, where "FDE" carries faint consulting connotations. A new title clears that association.

The result: the same role, in the same city, at the same seniority, listed under two different keywords, with a $127K spread between the top of one band and the middle of the other.

The talent gap everyone is blaming on supply is half a supply problem and half a Boolean problem.

How to source FDE talent when the title is unstable

Stop searching for a title. Search for a job description. The candidates you want share a specific shape of work history, not a specific string.

The pattern that generalizes across FDE, Agent Engineer, AI Application Engineer, and technical consultant is:

• Full-stack shipping experience (backend and at least one frontend framework), typically 4+ years.
• Customer-facing deployment work, either at a startup that sold to enterprise or a consultancy.
• Recent AI application layer work: LangChain, LlamaIndex, agent orchestration, RAG, or an in-house equivalent.
• Comfort in ambiguous scoping. Their commit history shows repo-hopping and infra work, not narrow feature ownership.

That set of signals is exactly what plain-English sourcing is built for, and it is the specific friction Refolk closes: you describe the person by shape of work rather than by title, and Refolk pulls candidates from GitHub, LinkedIn, and the open web whose profile matches, regardless of whether their current employer decided to call them an FDE, an Agent Engineer, or a "Member of Technical Staff, Deployment."

Consulting is the FDE market's dark matter

The largest FDE-shaped employer is not an AI lab; it is consulting, and most tech sourcers ignore it entirely. QuantumBlack (McKinsey) requires 8+ years of hands-on software, platform, or infrastructure experience for its senior consultant tracks. BCG's technical consultants ship agents at Fortune 100 clients. These are the same humans OpenAI's new deployment subsidiary wants.

OpenAI made this structural in May. On May 11, 2026, OpenAI launched the OpenAI Deployment Company, a $4B subsidiary built entirely around FDEs, and seeded it by acquiring a 150-FDE consulting firm. That acquisition, not any single Sierra job posting, is the clearest signal that consulting is now a farm system for AI deployment engineering.

If you are hiring FDEs and you are not sourcing from QuantumBlack and BCG, you are ignoring a large cohort of shape-appropriate candidates. The candidates are there. Their titles just say "Associate Partner" or "Senior Manager."

Sourcegraph's HN post is the tell

The Sourcegraph August 4, 2026 "Who is Hiring?" post collapses three arguments into one artifact. Sourcegraph is a company that hires FDEs, that also lists Agent Engineer as a distinct role in the same post, and that is doing so in the highest-signal recruiting channel for AI-native engineers. That combination tells you the fragmentation is now normal enough to appear in a single company's org chart, not just across companies.

Plank's census confirms this at scale. The exact-phrase "Forward Deployed Engineer" now appears in reqs at 569 companies across 22 verticals, with 1,031 live worldwide (1,391 all-time). If FDE postings and Agent Engineer postings were consolidated, the open-req count would climb meaningfully and the "no candidates" narrative would soften.

18x

The specific Boolean fix

Replace "forward deployed engineer" with a union query and add employer scoping. The generic upgrade looks like this:

("forward deployed engineer" OR "agent engineer" OR "AI application engineer" OR "agentic engineer" OR "software engineer, agent" OR "deployment engineer") AND (Palantir OR Sierra OR Vapi OR "Henry AI" OR Modal OR Roboflow OR Cresta OR "Gecko Robotics" OR "Amp Code" OR QuantumBlack OR "OpenAI Deployment")

That query still misses the consultants who describe themselves as "Senior Manager, AI" at BCG, and it misses candidates who never bothered to update LinkedIn. Which is why title-based sourcing has a ceiling.

The mechanism worth naming: on Paraform, 59% of FDE-hiring companies are Seed to Series A, 27% are AI-native by name, and 35% are Series B+. Early-stage AI-native founders invent new titles as a hiring wedge. That means title fragmentation is going to get worse before it gets better. Any sourcing stack that depends on exact-title Boolean will lose ground every quarter as new startups launch with net-new labels.

Description-based sourcing does not have that decay. When you tell Refolk "engineers with 4+ years of customer-facing deployment work who now ship AI agents," the query survives the next round of title inflation because it never depended on the title in the first place.

FAQ

Is "Agent Engineer" actually the same job as Forward Deployed Engineer?

At Sierra, Vapi, and Henry AI, yes, functionally identical. The job description (partner directly with customers, build and ship AI agents, engage with all levels of the customer's business) maps onto Palantir's original FDE definition. The pay is higher (Sierra's Staff Agent Engineer TLM hits $310K median base versus a $183K FDE midpoint on Paraform), and the cultural framing is different, but the daily work is the same. Sierra publicly runs both tracks in parallel, which is the clearest confirmation you can get from a single employer.

How big is the actual candidate pool if you union the titles?

Roughly 2,800 in the US, per Refolk's index: 2,654 titled "Forward Deployed Engineer" plus 148 with the fragmented titles. That excludes consultants at QuantumBlack, BCG, and similar firms who do the same work under non-engineering titles. The "no candidates" narrative comes from measuring only the first bucket against 5,330 April 2026 postings.

Why did AI-native startups reject the FDE title?

Three reasons. First, scarcity signaling: a new title supports a higher pay band. Second, brand distance from Palantir's consulting-adjacent connotation, which matters when recruiting from OpenAI and Anthropic. Third, product framing: startups selling "AI agents" want their engineering title to include the word "Agent" for narrative consistency with the product. None of those reasons will reverse in 2026, so expect more fragmentation, not consolidation.

What is the single highest-leverage change to make this week?

Rewrite every FDE req and every FDE search string to include Agent Engineer, AI Application Engineer, and Agentic Engineer as accepted variants. Add employer scoping to Sierra, Vapi, Henry AI, and Modal. If your sourcing tool can only Boolean on titles, either add description-based sourcing to the stack or accept that you will miss the fragmented cohort concentrated at the four employers your hiring manager most wants to poach from.