Turbopuffer first, LiteLLM as the fallback, Ploy only on exceptional terms
Asked:
“should i work at https://www.litellm.ai/ as their 4th enterprise sales rep in the USA, or turbopuffer.com as another enterprise rep, or https://ploy.ai/ as their 2nd rep overall? i want to maximize for earnings (w2 and equity combined). do a deep in-depth analysis.”
Three private AI-infrastructure companies, compared on public evidence current to 2026-09-07: 17 extracted source rows consolidated to one row per company, with 2–9 supporting URLs and 2–7 independent hosts each. Compensation listings are partly from aggregators, revenue is private and sometimes estimated, and undisclosed fields stay undisclosed — not inferred.
Decision matrix — analyst judgment scored 1–5 from the sourced evidence
Darker green = better for the candidate. Scores are analyst judgment built on the sourced rows below, not company disclosures. Hover a cell for the reasoning.
Company cards — disclosed facts vs undisclosed
What the matrix cannot say alone
Turbopuffer's role is closing inbound technical demand and POCs; Sacra estimates $100M annualized revenue in March 2026, up from $75M at 2025 year-end — treat as estimates, not audited financials.
sacra.comLiteLLM cites $10M+ ARR, profitability on $1.6M raised, and use by Adobe, Netflix and NASA; a prior founding-AE posting disclosed $100K–$175K base plus 0.05%–0.50% equity, while a newer AE listing shows $150K–$180K — as rep #4, do not assume the old equity range applies.
builtinboston.comPloy's founding-AE listing shows $250K OTE at a 14-person company with early customer traction, but no base/equity split and only two independent sources — the thinnest evidence of the three.
talents.vaia.comTurbopuffer's GTM has been primarily developer inbound via technical blog posts and a Cursor/Notion reference network — verify whether the reported $100M is GAAP revenue, ARR or annualized usage, and its top-5 customer concentration.
nextplayso.substack.com
How to value the offers
Expected combined annual value = expected cash compensation + probability-weighted, dilution-adjusted equity value ÷ expected tenure in years.
Private-company equity has no known value. Mandatory inputs before any grant can be priced: grant percentage, current fully diluted share count and strike price, last preferred price and post-money valuation, liquidation preference stack, refresh policy, vesting schedule, post-termination exercise window, and likely future dilution.
Pure arithmetic, not valuation claims: 0.10% ownership corresponds to $1M gross at a $1B common-equity exit before dilution, strike and tax; 0.50% corresponds to $750K gross at a $150M exit before those same adjustments. Grant size and exit size trade off — a bigger slice of a smaller likely outcome can beat a sliver of a giant one.
OTE is not expected W-2. It becomes expected W-2 only when territory, quota, live pipeline, ramp terms, crediting rules, accelerators and clawbacks support attainment — ask each company for 2025 and trailing-12-month attainment by rep, rep-level pipeline coverage, inbound share, median ACV, win rate, cycle length, named-account rules, expansion/renewal credit, payment timing, and caps.
Sign only if
- Turbopuffer: written offer of roughly $250K+ OTE, a realistic ramp or draw, a meaningful option grant disclosed as fully diluted percentage, and clarity on whether the reported $100M is GAAP revenue, ARR or annualized usage — plus top-5 customer concentration.
- LiteLLM: at least about 0.15% fully diluted, a credible named territory now that the role is rep #4, and evidence that open source drives paid conversion rather than self-hosting pressure.
- Ploy: roughly 0.5%+ fully diluted, $250K+ OTE, at least six months of non-recoverable ramp protection, and CRM/bank-statement evidence that the first rep and founders have a repeatable pipeline; confirm you are looking at ploy.ai, not the unrelated UK identity-security company that shares the name.
The rows behind the analysis
Method: one consolidated comparison row per company from 17 relevant extracted source rows (job listings, third-party revenue estimates, company pages), current to 2026-09-07. Matrix scores 1–5 are analyst judgment; disclosed comp and traction quotes come verbatim from the sources. Some listings are aggregators; company revenue is private and sometimes estimated; absent fields are shown as undisclosed, never inferred. Long evidence quotes trimmed in the table for space.