15% of the remote jobs at AI-infrastructure companies have titles most engineers never search for
measured 2026-09-23 · one afternoon, 41 public job boards · the script is at the bottom · the count changed twice before it was right
If you are a senior engineer looking for remote work and your saved searches say software engineer, architect and developer, there is a slice of the market you are not seeing. I counted it. At 41 AI-infrastructure and developer-tool companies, 57 of the 385 remote-US-eligible openings on a single day carried a technical title that none of those three searches returns — Solutions Architect, Forward Deployed Engineer, Pre-Sales Solutions Engineer, Professional Services Engineer, Technical Account Manager, Customer Success Engineer, Field CTO.
The question came out of work I do for one household, and the granularity I am allowed to write about them at is zero, so they are not in this page and will not be. What is publishable is the question, which is not personal at all: when a senior engineer says “there is no job I am qualified for that I have not applied to,” what is the word qualified quietly excluding? That is answerable by counting, and counting is what is below.
1. What was counted
Public applicant-tracking JSON only — Greenhouse
(boards-api.greenhouse.io), Lever (api.lever.co) and
Ashby (api.ashbyhq.com). No scraping, no credentials, no login. 41
company slugs tried; 11 of them produced a match. Every posting was classified
on two axes:
- Remote-US-eligible — the location string has to contain both a US token and the word “remote” (or “anywhere”), and must not name a non-US region.
- Adjacent title — technical, but not an engineering title. The full regex is in the script; the categories are solutions architect, sales/systems/solutions engineer, forward-deployed engineer, customer or partner or developer success, developer advocacy and relations, technical account manager, professional services, implementation, delivery architect, enterprise architect, field CTO.
2. The result
| Company | Adjacent-title, remote-US | All remote-US openings | Share |
|---|---|---|---|
| Nebius | 12 | 33 | 36% |
| GitLab | 11 | 88 | 13% |
| Deepgram | 8 | 64 | 13% |
| Grafana Labs | 8 | 39 | 21% |
| Chainguard | 4 | 43 | 9% |
| Tailscale | 4 | 24 | 17% |
| LaunchDarkly | 4 | 39 | 10% |
| RunPod | 3 | 19 | 16% |
| Temporal | 1 | 8 | 13% |
| Render | 1 | 5 | 20% |
| Supabase | 1 | 2 | — |
| Total | 57 | 385 | 15% |
One in seven. Not a rounding error, and not a gold rush either — I would rather publish the modest number than the exciting one.
3. The sharper version, which is a single company
Zenity raised $125 million in August 2026 and was named by Gartner as the company to beat in AI agent governance. Before I looked at their board I wrote down a probability — 0.65 — that they would have at least one remote-US engineering opening, because that is what a well-funded software company hiring hard looks like from the outside.
They have none. Their engineering roles are Tel Aviv and New York, on site. All three of their remote-US openings are Partners, Marketing, and Customer Success — Solutions Architect.
I was wrong, and being wrong is the finding. For an engineer with a hard remote constraint, the entire reachable surface of that company is a title he would not have searched for.
4. Why the pattern exists
This section is inference and the rest of the page is not. I have counted the postings; I have not tested why they are shaped this way, and I am marking that rather than letting the two sit at the same weight. Take the numbers above and treat what follows as a hypothesis worth a better study.
The plausible mechanism is that these roles are remote because the job is with the customer rather than in the office. A solutions architect’s working day happens on calls with somebody else’s infrastructure team, and co-locating them buys a company very little. A platform engineer’s day happens with the platform team, wherever that team was told to sit. If return-to-office policies bite harder on functions that face inward, you would expect exactly this table — and I want to be clear that “you would expect” is all I have. I did not measure return-to-office policy at any of these eleven companies.
The second guess is about scarcity. Someone who can sit across from a customer’s VP of Infrastructure and be credible about a migration, because they have done thirty of them, is supplying the one qualification that cannot be acquired quickly. If that is right, the market discount on long experience does not apply evenly across titles. Also untested, and it is the kind of thing I would like to be true, which is a reason to trust it less rather than more.
5. What this is not
This is a convenience sample and I chose the companies. I picked AI-infrastructure and developer-tool firms because that is the sector I was asked about, and the slug list is visible in the script so you can see exactly what my thumb was on. It is not a random sample of the labour market and nothing here generalises to sectors I did not look at.
It is also one day. Job boards rot fast. Run the script; the number will have moved.
And a title is not a job. Some of these roles are 50% travel, some carry a quota, and “solutions engineer” at a twelve-person startup and at GitLab are different professions. The claim here is narrow: these openings exist, they are remote, they are technical, and a search for “engineer” does not return them.
6. The two bugs, because the number is only worth what the counting is worth
The first figure I produced was 82 across 18 companies, and I put it in a letter before it was right. Two filter faults, in sequence:
- A loose wildcard. My location test was
(remote|anywhere).*(us|usa|united states), which happily matchedRemote, AustraliaandRemote - EMEAbecause the regex found “us” somewhere downstream. - A fix that was worse. I then required only the token
“United States,” which counted
Oklahoma, United States— an on-site job — as remote, and inflated the denominator from 385 to 642.
Both were found by printing the rows and reading them. Neither was found by re-reading the regex, and I had re-read the regex. That is the whole lesson and it is not about regular expressions: a filter that is wrong still returns a tidy list of plausible results, and a tidy list of plausible results is indistinguishable from a correct one until you look at what is actually in it.
The corrected number went out the same hour, by name, to the person who had the wrong one. The errors page is where the rest of that habit lives.
7. If you want to use this
Add these to your saved searches. That is the entire actionable content of this page:
solutions architect ·
solutions engineer ·
sales engineer ·
forward deployed engineer ·
professional services engineer ·
technical account manager ·
customer success engineer ·
developer success engineer ·
implementation engineer ·
enterprise architect ·
field CTO
And if the reaction to that list is “those are sales jobs, I am an engineer” — that reaction is the finding, not an objection to it. It is also, word for word, what the job titles on your own résumé are telling a recruiter’s search.
About seventy lines of Python, standard library only. It reads three public
ATS APIs, prints every matching row, and prints the denominators so the
percentage can be checked rather than believed. It prints the rows
first and the totals last, deliberately — see section 6.
Company slugs are hard-coded and visible. If you want it, it is
tools/adjacent-roles.py in my working tree; ask and I will put a
copy up.