Executive summary
Hybrid AI/ML roles pay a $44k premium over the remote + onsite weighted average ($267,260 vs $222,564). Remote averages $225,168. Onsite averages $221,158. The most surprising line in the data isn't the remote-vs-onsite pay gap — it's that the middle-ground arrangement, which many companies pitched as a compromise, is where compensation actually lands highest.
Top-line workplace mix
Across the 8,893 AI/ML engineering postings with a workplace classification, onsite is still the dominant mode — not by a small margin. Hybrid is the smallest bucket but commands the highest average advertised salary.
| Workplace | Role count | Share | Avg salary |
|---|---|---|---|
| Onsite | 4,748 | 53.4% | $221,158 |
| Remote | 2,564 | 28.8% | $225,168 |
| Hybrid | 1,581 | 17.8% | $267,260 |
The hybrid premium
Why does the middle option pay the most? Two factors explain most of the premium. First, seniority. Hybrid roles skew toward staff, lead, and principal bands — the levels where companies want engineers in the office for architecture discussions and design reviews but accept that 30+ year-old senior engineers won't relocate full-time. The salary endpoint reports a single weighted average of $231,380 across all salary-disclosed roles, and the hybrid band sits $+35k above that global average — that's not a remote-work discount, it's a seniority mix.
Second, geography. Hybrid postings anchor to a specific metro (usually San Francisco, New York, or Seattle) and inherit that metro's pay band. Remote postings in the same index compete with distributed talent from lower-cost geographies and from candidates who've already priced remote work into their expectations. The $225,168 remote average still out-earns most non-AI engineering roles, but it reflects a broader supply pool than the $267,260 hybrid average does. The practical read: hybrid is where AI companies are paying SF+NYC premiums plus demanding most-week attendance. It's the most expensive arrangement to hire into because it selects the most constrained candidate pool.
The industry lore says remote kills the pay premium. The live index says otherwise: the middle-ground workplace arrangement wins on absolute dollars, and it wins because it concentrates SF/NYC metros and staff+ seniority into the same candidate pool.
Onsite-heavy companies (80%+ onsite)
The following companies ran 10 or more AI/ML postings in the live index with at least 80% marked onsite. This list skews toward robotics, autonomous systems, and frontier labs — all areas where physical hardware, secure facilities, or classified work makes remote collaboration operationally difficult. It explains the 55% onsite floor in the top-line mix.
| Company | AI/ML roles (sample) | Onsite | Onsite share |
|---|---|---|---|
| Scale AI | 47 | 47 | 100% |
| Databricks | 42 | 42 | 100% |
| Glean | 32 | 32 | 100% |
| Lila Sciences | 26 | 26 | 100% |
| Thinking Machines | 23 | 23 | 100% |
| Together AI | 18 | 18 | 100% |
| CoreWeave | 18 | 18 | 100% |
| C3 AI | 18 | 18 | 100% |
| Roblox | 16 | 16 | 100% |
| DoorDash | 15 | 15 | 100% |
Remote-friendly companies (50%+ remote)
Conversely, here are the large-sample companies leaning remote-friendly (50%+ of AI/ML postings marked remote). These are typically distributed developer-tool companies, AI infra, and companies built remote-first from day one — they tend to set a remote salary floor at or above major-metro onsite pay.
| Company | AI/ML roles (sample) | Remote | Remote share |
|---|---|---|---|
| Instacart | 13 | 13 | 100% |
| 27 | 24 | 89% | |
| Harvey AI | 20 | 17 | 85% |
| Torc Robotics | 12 | 8 | 67% |
| 27 | 16 | 59% | |
| Nebius | 18 | 10 | 56% |
| May Mobility | 17 | 9 | 53% |
Salary by workplace
A simple horizontal bar chart of average advertised salary by workplace mode. Hybrid leads, remote and onsite are within a few hundred dollars of each other.
Regional bias
Regional bias caveat. The AI Dev Jobs index over-represents US-based postings. US pay-transparency laws (California, New York, Washington, Colorado) drive salary disclosure, so US postings are weighted more heavily in the salary averages than in the posting counts. European and APAC AI/ML roles are present in the index but many publish salary as "competitive" and drop out of the $3,482-posting salary-disclosed subset. Remote postings in particular often advertise a US-only salary range even when they're open globally — interpret the $ figures as US-anchored.
Practical takeaway
If you're choosing what to apply for. The headline-grabbing remote-vs-onsite framing is the wrong axis. Hybrid is the highest-paying band because it picks up SF/NYC salary floors plus senior/staff+ level mix. If you have the flexibility to be in-office 2-3 days a week in a major metro, hybrid is where the dollars land. If you need fully remote, target the remote-friendly companies in the table above — they set the remote floor higher than the $225,168 index average. If you're early-career, onsite roles at robotics, autonomous-vehicle, and frontier labs are where the 53% onsite share concentrates — lean in, physical-AI work is where onsite still pays real premiums.
Methodology
Generated from the live aidevboard.com/api/v1/stats endpoint (public, unauthenticated) plus a paginated walk of /api/v1/jobs for per-company workplace aggregation. Index scrapes 560+ ATS sources on a daily cron and deduplicates by (company, title, location). The workplace field is classified by a rules-based parser from job-title + description + location signals; edge cases ("hybrid, 3 days", "flexible", "in-person preferred") are normalized into the three canonical buckets. Of 8,893 roles in the index, 8,893 have a non-empty workplace classification; the rest are silently excluded from this analysis. Salary averages use employer-advertised midpoints from the 3,482-posting salary-disclosed subset. This page auto-regenerates weekly (Mon 9:00 am PT).
Download raw data: The workplace-by-salary dataset is mirrored as a public gist — CSV · Markdown · view on GitHub. Auto-updated every weekly regeneration; canonical raw URLs are stable across revisions.
What's next
For the skill-and-compensation view of this same dataset — which tags actually pay the premium — see Q2 2026 AI Engineering Compensation by Skill. For the company-side view of where the roles are concentrated, see Q2 2026 AI Engineering Hiring Snapshot. For the infrastructure side — how many of the agent systems these engineers are building actually have a working MCP endpoint — see Q2 2026 MCP Ecosystem Health. Full reading paths at the Research Atlas.