The evidence behind the figures
Useful context. Clear limits.
What this guide is
- Selected public employer salary disclosures.
- Source-linked examples in their advertised currency.
- Context for a more specific compensation conversation.
What it isn’t
- A representative survey or market average.
- A record of accepted offers or live vacancies.
- A recommended package or guaranteed equity value.
Modelled estimates: version 2
Baseline: well-funded, venture-backed startups and scaleups at Series A/B. Use the modifiers below for stage and company type.
The US model is built from 1,046 advertised ranges at 220 startups, collected on 24 October 2026, excluding public companies, big tech and frontier labs. Modelled values never enter real advert counts or observed medians. At least 10 matching adverts are required before observed figures replace the estimate.
London uses Ravio’s published UK growth-stage software-engineering benchmarks, with a 15% uplift chosen for Corellia’s premium, well-funded target market. That uplift is a modelling assumption, not a measured percentile or an observed employer offer. Mid-level and Senior are anchored separately; Staff, Lead and Founding use US-derived level ratios.
What moves this number
The supplied model applies specialism adjustments of 8% for AI and machine learning, 25% for Research and 16% for Technical leadership. Funding-stage adjustments vary by level, with more modest base-salary increases after Series B. These are model inputs, not premiums measured in every selection.
The separate frontier-lab or AI-native scenario applies a 1.35 multiplier. Its evidence and experience-based assumptions are described in the full method. It does not classify or alter the observed advert cohorts.
Modelled ranges are approximately 13% either side of the median, rounded to the nearest 1,000. Equity and bonus guidance stays separate from salary and never becomes a guaranteed package. The UK and US level definitions are not directly equivalent, and the UK evidence is thinner.
Read the full supplied method and modifier tables, 24 October 2026Read the full methodology
This edition contains 3924 role-level examples from 308 employers. The latest additions were checked on . There are 3184 distinct source adverts. Some adverts contain multiple levels or city ranges; those are shown separately without being counted as independent adverts. These are not independent samples of the whole market.
The expansion reviews public employer careers pages and YC company listings. It is a convenience sample, not a random or representative survey. The original expansion capped additions at 30 per employer and 20 for specialist comparators OpenAI and Applied Intuition. The 23 October update adds public employer ATS disclosures, preserving explicitly named secondary locations and city-specific compensation tiers. Multiple adverts from one employer are not independent market observations; the employer count is shown alongside n. We selected roles relevant to technology startup hiring and retained examples where a salary disclosure could be verified directly. We excluded unavailable or ambiguous disclosures. Advertised roles may close or change after the review date; this page is salary research, not a live job board.
The primary view matches city, specialism and explicit level, giving 60 combinations across three cities, four specialisms and five levels. Funding stage is available for every selection. Advert cohorts with at least 10 records use observed percentiles; smaller cohorts use the separately labelled modelled estimates described below. Adverts and official LCA filings are counted and summarised separately, including when All is selected. LCA wages are employer-offered amounts, not proof of actual pay. Unknown levels and other Bay Area cities do not enter named San Francisco cohorts.
Within each source type we use an explicitly supplied salary point, or the arithmetic midpoint of the stated range, and nearest-rank percentiles without interpolation. At least 10 records are required for observed percentiles. Smaller advert groups use modelled estimates, while smaller official-filing groups show observed bounds only. Repeated source/case identifiers in the same city and level count once. Published reports remain a separate panel; report-wide samples are never used as cohort counts. Equity and bonus remain separate disclosures, never assumed to be zero or combined into guaranteed compensation. Observed figures use no currency conversion or inferred funding-stage premium. Modelled estimates use the supplied regression adjustments, separately from observed figures.
Staff/Principal combines those two levels. Lead/Head of groups technical leads and engineering management, including heads, directors, VPs and CTOs; these scopes differ and require individual calibration. Series C+ includes disclosed private-company rounds from Series C onwards; the exact round remains in each source record. Unspecified levels and stages never enter a named cohort. Numeric equity needs its own five-source threshold. When a selected role has fewer than five numeric disclosures, the component can use the same city and funding stage across all technical specialisms and levels, clearly labelled. Equity and bonus availability counts use only explicit disclosures, never an assumption about unmentioned benefits. Bonus targets, maximums and fixed awards are not pooled into a typical percentage. We do not infer missing equity, bonus or a typical total package.
Review dates describe the source checks, not the date this tool changed. Sources are reviewed manually; there is no automatic live salary feed or promised update frequency. Every selected view is shareable using its query string, and the browser remembers the last selection in first-party preference storage.
Coverage includes junior, mid, founding, senior, staff, principal and engineering leadership roles, with named funding stages from pre-seed to later private-company rounds. 1362 adverts have no verified funding-stage label and are excluded when you select a named stage. Headcount is total company size where stated, not engineering hire order. Pre-seed, VP-level and some city-by-role combinations remain thin or absent. We do not establish complete market bands, contract rates or a reliable equity or bonus benchmark for every stage.
Funding stages are taken from linked source statements, not inferred from investment amounts, accelerator batches or a company’s customers. Stage labels describe the reviewed source and may lag a subsequent funding round. A funding-announcement date is shown when verified; where the advert itself states the round, its date records that stage evidence, not the date the round closed. Equity amounts are grants as advertised, not annualised cash or guaranteed proceeds. Undisclosed bonuses and equity are not zero. Disclosures with conflicting salary figures are excluded.
0 retained research records are excluded from the displayed figures because their sources are unavailable, hiring is paused, adverts overlap with differing terms, or the location is outside the selected city. Each source has a review date. Review dates record when we checked a disclosure; they do not mean the advert is still live today. If a figure cannot be reverified, it will be removed or clearly identified as historical rather than silently treated as current.
Suggest a correction