Back to Career Hacks

We Rescued Salary Data From 22,000 Startup Job Postings. The 'Series D Trap' Was An Averaging Artifact.

By Fast AI Startup Jobs

Every "startup salary" article you've read shares the same blind spot. The structured compensation field in most ATS systems is filled in for almost nobody — in our dataset of 35,574 raw postings, exactly three had a salary listed in that field. So most analyses either ignore comp entirely, or back-fill it from self-reported sources like Levels.fyi (sparse and self-selected) and pretend that's the answer.

But job postings have bodies, not just structured fields. California's SB 1162, in effect since 2023, forces any company with 15+ employees to disclose a pay range inside the posting itself if the role can be performed in CA. We walked every cached JD we had — 22,529 of them — with a regex for $X – $Y ranges, sanity-bounded the values, classified each as base / OTE / unspecified, and recovered 4,437 structured salary records. That's a real 19.7% job-level coverage, ~2,000× more than what ATS metadata alone offers.

Once we had that, we cross-joined with the funding stage, headcount, and industry of each company. Three of the findings we had previously published — based on hiring patterns alone — partially broke. Here's what the data actually says.

Open the full interactive atlas — 18 charts, hover for company-level detail →

All charts referenced below are interactive in the atlas. This article gives you the headline numbers; the atlas gives you the company-level lookups.


TL;DR — 10 things this data changed our mind about

  • The Series D "Sales gold rush" is real on average but false in three industries. FinTech sales hiring peaks at Series A–B (29%) and falls by D. Enterprise SaaS and DevTools peak at Series C. The "D trap" framing only cleanly fits Cybersecurity, AI/ML, and Hardware/Climate.
  • Series A is the lowest-paid well-sampled stage. Median base $150k — lower than Seed.
  • "Seed" stage is now two completely different job markets. Traditional Seed pays $100–150k base. The new cohort of frontier AI labs raising $50M–$300M+ at "Seed" pays $250–450k. Ask about last round amount before evaluating any "Seed" offer.
  • AI Infrastructure / Models is the highest-paying industry at every stage. Series A AI Infra ($210k median base) pays more than Series D FinTech.
  • Stage alone is a weak predictor of pay. Industry × stage is far stronger.
  • GTM jobs in postings that explicitly mention "OTE" cluster ~20% higher than those quoting base. Most companies disclose one or the other, not both.
  • Engineering base pay is 20–40% above GTM / Operations at every stage. Even at Series D when GTM share of headcount is highest.
  • By Series D, nearly a third of companies haven't closed a round in 24+ months. The "stage" label hides this entirely.
  • "Founding X" titles vanish after Series A (4.4% of Seed postings → 0.1% of Series D).
  • Series E is "Staff Engineer inflation" season. "Staff" appears in 19% of E-stage titles, 3× the rate at Series B.

How we got the salary data

Most ATS feeds expose a "compensation" field. In our crawler outputs across Greenhouse, Lever, Ashby, Workable, Comeet, Gem, and Consider, that field was empty or "Not disclosed" for 99.99% of postings. Three jobs had structured comp out of 35,574.

The postings themselves, however, often include a salary range in the body — particularly for roles that can be performed in California, where SB 1162 mandates disclosure. So we wrote a multi-pattern regex extractor: matched $X – $Y patterns (handling hyphens, en-dashes, em-dashes, "to" connectors, and "K"-suffix shorthand), sanity-bounded the values (annual $20k–$1.5M, max:min ratio under 3.5×), and classified each match as base / OTE / unspecified based on contextual keywords ("base salary", "OTE", "total compensation").

Yield: 4,437 records from 22,529 cached JDs (19.7%). After deduplication and joining to company stage and headcount data, 4,176 records survived with all three dimensions — about 18.6% of our unique posting universe. Coverage skews toward California-based roles. Salary charts in this article use base only, since mixing base and OTE inflates GTM medians artificially.


The mapping: median base pay by funding stage

Base pay only, USD, midpoint of disclosed range:

StageNp25Medianp75
Seed36$165k$200k ⚠️$250k
Series A230$125k$150k$191k
Series B342$158k$190k$215k
Series C317$160k$193k$235k
Series D183$145k$170k$198k
Series E120$140k$168k$225k
Series F153$169k$203k$230k
Series G+58$115k$140k ⚠️$179k
Private Equity59$163k$200k$235k

Two cells deserve a flag. The Seed median ($200k) is higher than every nearby stage — that's not real, it's a selection effect. Companies disclosing salary at Seed are dominated by well-funded AI labs paying frontier-research premium. We dig into this below. The Series G+ median ($140k) is the lowest of any well-covered stage — also not because late-stage companies underpay, but because mature companies post many more mid-level / IC roles publicly while paying their senior staff $300k+ behind the curtain.

The full Stage × Department mapping (Engineering vs Sales/GTM vs Operations vs Other) is in the interactive atlas. The richest cell across the entire mapping is Series C and F Engineering at ~$214k median base. The leanest cell with a defensible sample is Series A Operations at $140k.


The "Series D Sales gold rush" doesn't hold by industry

We previously argued that Series D was the peak GTM-hiring stage — sales as % of postings hits its highest point there. That was the pooled finding across all industries.

When we cut by industry, the picture fractures:

IndustryWhere GTM share actually peaks
CybersecuritySeries D (43%) ✓ matches the pooled story
AI / MLSeries D (30%) ✓ matches
Hardware / Climate / DefenseSeries D (21%) ✓ matches
Enterprise SaaSSeries C (41%) — peaks one stage earlier
DevTools / Data InfraSeries C (33%) — peaks one stage earlier
Healthcare / BioSeries B (27%)
FinTech / CryptoSeries A–B (29%) — falls through Series D
Consumer / Media / GamingSeries C (28%)

The "Series D Trap" framing is correct for AI, Cybersecurity, and Hardware — but for FinTech and SaaS, you're targeting the wrong stage by a full round or two. If you're a salesperson optimizing for "where am I most wanted right now," your answer depends on which industry you've already chosen.

This is precisely the kind of finding that gets obscured when 8 different industry curves get smashed together into one "all-stages-averaged" view. We should have done this cut from day one.


"Seed" is now two completely different job markets

The Seed-stage salary median ($200k base) being higher than Series A ($150k) wasn't an extraction artifact. We audited every Seed company with a disclosed base salary:

CompanyFunding raised at "Seed"Disclosed baseRole
Periodic Labs$300M$350k–$450kSupercompute Engineer
H Company$220M$110k–$140kDemand Generation Manager
Collate$30M$150k–$350kAI Engineer / Researcher
RunPod$22M$200k–$320kPrincipal Security Engineer
Inference$11.8M$250k–$350kML Researcher
Bayesian Health$15M$145k–$185kProduct Manager
Arlo$4M$90k–$110kSenior Member Support

The "Seed" label is no longer load-bearing. A Seed company that raised $5M and a Seed company that raised $300M are running entirely different businesses, hiring entirely different people, and paying differently — sometimes by a 3× factor.

Before accepting any "Seed" offer in 2026, ask the funding amount, not the stage. A "$300M Seed" company has functionally already raised a Series C — they just opted to keep the Seed label, often to avoid the round-stacking dynamics that bigger-name stages create.


The companies that should make you ask about runway

We took every company with (a) more than 18 months since their last funding round and (b) currently open jobs equal to 5%+ of their headcount and called them "stale + still hiring hard." 50 companies fit the pattern. The top of the list:

CompanyStageDays since last roundOpen jobsHiring intensity
VarianceSeed2,532 (≈ 7 yrs)132.60
Bayesian HealthSeed1,772 (≈ 5 yrs)120.40
PlaygroundSeries A1,050 (≈ 3 yrs)200.67
Aqua VoicePre-Seed815 (≈ 2 yrs)71.40
IntroSeries A715 (≈ 2 yrs)81.60

A "stale + still hiring" company can mean one of three things: (a) they're bootstrapped to profitability and don't need to raise — good; (b) they're sprinting to revenue before a Series X — neutral, ask; or (c) they're about to hit a wall — bad. The data alone can't distinguish them. What it can tell you is which question to ask in the final-round interview. If you're interviewing at any of the 50 companies in the interactive atlas's red-flag chart, runway is the question.


How to use this when you're job hunting

A condensed decision matrix. Each row maps a goal to the stage and industry combination the data actually supports — not the conventional wisdom version:

If you want…Target stage + industryTypical baseWhy the data says so
Highest cash baseAI / ML, Series C~$275kThe richest cell in the full mapping
Founding-title roleSeed → early A (verify funding < $30M)$150–250k"Founding" 4.4% of Seed → 0.1% by D
Sales gold rush (AI sector)Series D AI / ML~$155k base (OTE 20% higher)GTM share 30%, highest of well-sampled AI cells
Sales gold rush (FinTech)Series A or B FinTech — not D$130–170k baseGTM peaks at A in this industry, falls through D
Frontier-AI research role"Seed" companies w/ ≥ $50M funding$250–450kThe new-Seed cohort pays Series E rates
Staff Engineer titleSeries E~$188k"Staff" appears in 19% of titles, 3× Series B
Remote-friendly roleSeries C–E$180–195k baseRemote share peaks ~20% in this band
Career stability over upsideSeries F+ (verify last round < 24 mo)$140–210kConfirm with the red-flag check first

What this analysis is not

A representative survey of the US startup market. Our tracked company list is curated for AI & tech relevance. The salary extraction skews toward California-based postings (where SB 1162 forces disclosure). The data is a single-point snapshot, not a flow — when we say "Series D FinTech peaks at GTM," we mean reqs currently posted; the answer 6 months ago or 6 months from now could differ.

Equally important: we make no statistical inference claims. Every comparison above is a point estimate. We don't yet have bootstrap confidence intervals on these medians, and we don't yet have a role-level salary breakdown (a "Senior SWE" and a "Staff ML Researcher" both fall into the "Engineering" bucket but pay very differently). That's the next layer of analysis. For now, take these numbers as directionally correct, not as a guarantee.

Open the full interactive atlas with all 18 charts →

Hover any chart for company-level data. Cell-level sample sizes are visible on the heatmaps.

Analysis built from a snapshot of 22,403 unique job postings across 887 actively-hiring AI & tech startups, plus 4,437 base-salary records extracted from JD bodies. Built May 2026.