The Database Behind Billion-Dollar Bets
Engineering challenges at the intersection of finance, law, and distributed systems.
Feb 18, 2026 — 4 min read
Written by
Every breakout startup works hard to get the right investors on its cap table. Those venture funds work just as hard behind the scenes: raising capital from limited partners, allocating it with contractual precision, and tracking every mutation over the fund's lifetime. Venture is a database of promises that change over time, and our job at AngelList is to ensure those changes are legible, auditable, and automatable. Today, AngelList builds infrastructure for venture capital. We build the source of truth for funds' legal and economic reality, maintaining audit trails accountants can trust, and a product investors can rely on.
What is fund administration, anyway?
One of the categories we play in is fund administration: A venture fund raises money and writes checks, but in practice it is a long-lived contract that encodes:
- Who promised what (commitments)
- Who actually paid what (capital calls + cash transfers)
- What the fund bought (investment assets)
- How the fund pays for itself (fees + expenses)
- How upside is split (carry waterfalls)
All of that changes over time, often retroactively, because the real world is full of edge cases, and lawyers are endlessly creative.
AngelList is the fund administration platform behind a big chunk of venture’s plumbing. Fund managers using our software represent $171B in assets on our platform across ~25k investment vehicles. We are a small engineering team of less than 50 with absurd impact: more than half of top-tier U.S. VC deals involve investors on AngelList.
The engineering challenges are real
We model contracts as evolving state machines, reconcile noisy real-world inputs into canonical entities, and turn messy legal reality into append-only accounting facts. We have built primitives for backfills, reversals, and historical replays — so when lawyers and accountants change their understanding of the past, our systems don’t flinch. We're building distributed systems fundamentals applied to finance, where the packets are PDFs and the failures cost real money.
