Forward deployed engineering: the verified numbers

Almost every number circulating about this role is unsourced. Below is what we could verify, what it rests on, and what to distrust.

Demand and hiring

Postings up 1,165 percent year over year. Comparing January to October 2025 against the same months in 2024, measured by Revealera and published by Henley Wing Chiu on 18 November 2025. October 2025 was the highest posting volume on record in that dataset.

Caveat that matters: the same analysis found the title covers three different jobs, so part of the growth is adjacent roles being relabelled rather than new work appearing. Growth in a job title is not the same as growth in a job.

The three archetypes

From the same 1,000 postings:

• Builder, 60 percent. Production engineers embedded with customers. 70 to 90 percent of time coding. Advertised range $140,000 to $250,000.
• Sales engineer with implementation duties, 30 percent. 30 to 40 percent of time coding. Advertised range $120,000 to $200,000 plus commission.
• Internal tools builder, 10 percent. No external customer. Advertised range $100,000 to $180,000.

This is the single most useful statistic on the page, and it gets its own breakdown.

Compensation

Median advertised salary $173,816 across the 1,000 postings.

70 percent mentioned equity. 8 percent mentioned commission or on-target earnings. None were quota-carrying. That last figure is the cleanest test for separating the role from sales engineering.

Palantir forward deployed engineer total compensation: median $278,000, range $185,000 to $631,000. Self-reported to Levels.fyi, United States, read on 8 September 2026. Self-reported aggregator data skews toward people who choose to report, so treat it as indicative.

What we could not verify: any specific compensation band for OpenAI or Anthropic forward deployed engineers. Levels.fyi does not have enough submissions to show a level breakdown for OpenAI's forward deployed engineer title. Ranges for frontier labs appear in many blog posts and we could not trace one to a primary source, so we do not publish them.

Experience and background

Experience targeted, from the 1,000 postings: 12 percent at zero to two years, 60 percent at three to five, 20 percent at six to eight, 8 percent at nine or more. The role is staffed mid-level.

Routes in: 45 percent from software engineering, 22 percent from solutions or sales engineering, 15 percent from data engineering or data science, 10 percent from technical consulting, 8 percent founder or early employee.

Alongside that, First Round Review's February 2026 guidance reports that ten or more years at a large tech company can be a red flag for this work, because the role needs people who operate without institutional scaffolding.

Skills named in postings

Languages and platforms: Python 66 percent, TypeScript 35 percent, AWS 32 percent, GCP 22 percent, Azure 18 percent.

AI competencies: AI agents 35 percent, LLM experience 31 percent, RAG 12 percent.

Soft skills: customer communication 47 percent, followed by problem-solving under pressure and adaptability across industries.

Duties named most often: working directly with customers 55 percent, building and deploying AI or ML systems 37 percent, integrating systems and APIs 32 percent.

Travel

68 percent of postings required travel. Where a figure was given, 50 to 75 percent was typical.

OpenAI's own postings for forward deployed software engineers state that up to 50 percent travel is required. That is the employer's own statement about its own role, which makes it a primary source for that company.

Who is hiring, and for which customers

58 percent of postings came from companies with 11 to 200 employees. The model is being adopted by growth-stage companies selling into large enterprises, rather than only by the well-known names.

45 percent sat on a dedicated forward deployed team rather than reporting into sales. Where the function reports covers why that number matters more than it looks.

79 percent did not specify a target customer industry. Among the 21 percent that did: financial services and banking 24 percent, government and defence 18 percent, healthcare 17 percent, insurance 17 percent, energy and utilities 13 percent.

Investment

AWS committed $1 billion in June 2026 to a Forward Deployed Engineering organization embedding thousands of engineers with customers, announced on its own newsroom. It stated the work would be priced on shared goals and business results rather than billable hours, and named customer self-sufficiency after the engagement as a design goal.

OpenAI had 22 open forward deployed and solutions engineering roles out of 311 total positions at the time a16z's Trading Margin for Moat was published on 4 June 2025.

The margin numbers behind the argument

From the same a16z piece, on the previous generation of enterprise software: ServiceNow's gross margin at IPO was 63.2 percent, Workday's 54.1 percent. By 2024, ServiceNow was at 79 percent and Workday at 75 percent. The services-heavy start was the on-ramp rather than the business model.

The most-quoted number, and its sample

95 percent of generative AI pilots delivered no measurable profit-and-loss impact. MIT's Project NANDA, "The GenAI Divide: State of AI in Business 2025", July 2025. Only 5 percent of integrated systems created significant value.

The sample: 52 executive interviews, surveys of 153 leaders, and analysis of 300 public deployments. That is real research and it is not a census. Almost every article quoting this figure omits the sample size. Include it, and the number becomes a defensible signal rather than a slogan.

Numbers we deliberately do not publish

Stated plainly, because their absence is the point:

• Any frontier-lab salary band. Widely published, not traceable to a primary source.
• "Palantir employed more forward deployed engineers than product engineers by 2016." Repeated often, no primary source found.
• A benchmark ratio of forward deployed engineers to product engineers. No credible published figure exists. Our page on sizing gives the arithmetic instead.
• Any average engagement length or time to production. These vary with the environment and nobody has published a defensible distribution.

If you find a primary source for any of these, it belongs here. Until then, a page that says "unknown" is more useful than one that guesses.

How to cite this page

Every figure above links to its origin. Cite the origin rather than us: we are an index into other people's measurements, plus the caveats their coverage tends to drop.