AI Comp by Funding Stage.

AI compensation does not move linearly with seniority and city alone. It bends sharply with company funding stage. A seed-stage AI company in San Francisco can pay $165K base with a $95K-per-year equity grant for a senior AI engineer. A public AI company pays $268K base with $112K-per-year RSU. The total comp numbers happen to be in the same ballpark for the headline ($260K vs $408K when equity is liquid), but the composition is dramatically different and the risk profile is incompatible.

Without a stage-aware engine, every comp calculation has to assume an average company stage, which makes the answer wrong for both seed and public. The Comp by Stage engine fixes that by treating funding stage as a first-class dimension. Pass company_stage=series-b to /calculateand the response returns the band for that stage specifically. Without it the API uses the role's stage-weighted average and flags the assumption in the response.

We built this engine because every founder, every candidate evaluating multiple offers, every comp consultant building bands for an early-stage portfolio, and every recruiter writing offer narratives needs stage-specific data. The market did not have a clean source for this: Levels.fyi conflates stages within companies; Pave is paywalled and stops at seed-stage detail. Orbyt's engine surfaces the breakdown in every salary response, with sources cited and methodology version visible.

For agents, this engine is the difference between "an AI engineer earns $310K in San Francisco" and "a senior AI engineer at a Series B company in San Francisco earns $225K base + $58K/yr in equity + $15K signing, totaling $298K: versus $408K at a public company in the same city, with the gap explained almost entirely by the equity liquidity profile." The second answer is decision-ready; the first is a rough average that misleads both sides.

Stage is determined by tracking each company's most recent funding round. Signals come from the dataset's traceable sources: H-1B LCA filings (DOL), state pay-transparency job postings, and community submissions. We cross-reference the round size against the company's reported valuation to filter out misclassified rounds (a $50M Series A is sometimes mis-tagged as a Series B; we use the size to align with the round's actual stage).

Compensation is composed of four components: base salary (from H-1B LCA filings + job postings + community submissions), equity grant value (from the most recent priced round or trailing 30-day VWAP for public), vesting schedule (default 4-yr cliff-1 if not disclosed), and signing bonus (from community submissions + job-posting bonus disclosures). The engine returns each component separately so customers can apply their own valuation logic.

For private companies, the equity value is the face value at the most recent round. We never silently apply a liquidity discount: the customer applies their own. Some customers use 30-50% for Series A, 20-40% for Series B, 10-30% for Series C+; the appropriate discount depends on the customer's risk tolerance and the company-specific liquidity profile, neither of which we can know in aggregate.

The stage-specific multiplier on base is fit quarterly with the same mixed-effects model described on the Skill Premiums page. Stage enters the regression as a fixed effect with seven levels; the coefficients on each stage are the stage-specific premium / discount relative to the cross-stage average for the role.