In [4]: # tier_bands.ipynb
Six employer tiers capture the meaningful variance in ML engineer compensation. T1 is the frontier of the labour market, T6 the earliest-stage startup. Bands describe a generalist L5 senior; other levels scale per the level distribution.
T1
Foundation-model labs (post-training scale, 100B+ parameter models)
$220k - $480k
TC $500k - $2M+
Equity & bonus
Pre-IPO equity with 2024-26 valuations driving outsized paper value
Bonus: 10 - 20 percent of base
Notable trade-offs
In short
Highest TC ceiling in the industry; equity dominates
T2
Trillion-dollar platform companies with public RSUs
$185k - $300k
TC $280k - $700k
Equity & bonus
Public RSUs vesting over 4 years; predictable annual refreshes
Bonus: 15 - 25 percent of base, plus performance multiplier
Notable trade-offs
In short
Predictable comp progression with banded levels (L3 to L8)
T3
Series C - E private companies built around AI infrastructure or products
$170k - $260k
TC $260k - $520k
Equity & bonus
Pre-IPO equity with high variance in eventual outcome
Bonus: 10 - 15 percent of base
Notable trade-offs
In short
Higher upside than T2 if exit succeeds; 0.05-0.5 percent grants
T4
Systematic trading funds, market-makers, HFT
$200k - $350k
TC $350k - $1M+ cash
Equity & bonus
Profit-sharing bonus model; partner track at the top
Bonus: 50 - 200+ percent of base, performance-tied
Notable trade-offs
In short
Highest cash compensation; no equity but bonuses can be extraordinary
T5
Non-tech Fortune 500, healthcare, retail, large financial
$130k - $200k
TC $155k - $260k
Equity & bonus
Smaller RSU grants; sometimes none below senior level
Bonus: 10 - 15 percent of base
Notable trade-offs
In short
Stable compensation with broader scope per engineer
T6
Pre-Series B, seed-stage AI companies
$120k - $180k
TC Variable; often base + equity %
Equity & bonus
0.1 - 1.0 percent ownership; high variance outcome
Bonus: Often absent
Notable trade-offs
In short
Lowest base; equity is the lottery ticket
The tier bands above are the durable model. For the specific employers where demand and pay move fastest, these deep dives carry named figures re-checked regularly against Levels.fyi and company filings, mapped back to the tier they sit in.
OpenAI vs Anthropic ML engineer salary T1
Side-by-side total comp by level; PBC equity vs capped options; IPO and liquidity
Anthropic ML engineer salary T1
Single-lab deep dive: L4 to L7 bands, capped-option equity, the autumn-2026 IPO
Quant trading firm salary T4
Cash-bonus model at Renaissance, Two Sigma, Citadel, Jane Street, DE Shaw
FAANG ML engineer salary T2
Google DeepMind, Meta, Amazon Applied Scientist, Apple AI, Netflix bands
Frontier AI lab tier T1
T1 intake, equity forms, PhD vs MS, and the $500k-$2M+ ceiling explained
Non-FAANG ML engineer salary T1/T3/T4
Where frontier labs, quant, and AI unicorns beat big-tech total comp
Q.Why don't you list specific employer salaries?
▸A.The foundation-model labour market is re-pricing roles faster than salary tables can be maintained. A specific number attached to a named employer dates within weeks; an anonymised tier band remains useful as the market shifts. We use six tiers because they capture roughly 95 percent of the meaningful variance.
Q.Do quant trading firms pay more than big-tech in cash?
▸A.Yes, in cash compensation, T4 quant trading firms typically pay more than T2 hyperscalers at senior+ levels. Base salaries of $200k-$350k plus profit-sharing bonuses of 50-200 percent can exceed $1M cash for top performers. T2 hyperscalers offset with public RSU grants and predictable refresh cycles, which can add up to comparable or higher values over a 4-year stretch.
Q.Is it worth joining a T6 startup for the equity upside?
▸A.Sometimes. The maths: a 0.5 percent stake at a $500M exit yields $2.5M; most startups never reach that. Evaluate stage, market size, founder quality, and your personal risk tolerance. The expected value of T6 equity is usually negative; the variance is what makes it attractive.
In [ ]: # reader_questions.ipynb
Write in. A person opens every message, and the ones we can answer come back with the survey release or the public disclosure the number is drawn from, stated plainly enough that you can go and check it. We put no clock on a reply. Should an answer be worth reading twice we might set it out here stripped of anything identifying, and we will write and ask you before that happens.
Out[ ]:
Treat everything on this site as compensation reference assembled from published surveys and public filings, not as financial advice and not as a benchmark any single negotiation should be anchored to. Your own offer turns on the employer, the level it is pitched at, the equity instrument behind it and the market on the day it is made.