The archive · AI & Models · Product decision · 2025–2026
AfterQuery's expert-data bet: $30M Series A at $300M, $100M run rate in 15 months
Founded Jan 2025 to sell professionals' recorded reasoning to frontier labs; by April 2026 it had ~100,000 experts, $30M Series A and a $100M run rate.
AfterQuery
What the business is
Sells AI labs curated training datasets and practice environments built from ~100,000 verified professionals' recorded reasoning.
Starting capital：$30M Series A at $300M valuation (April 2026); prior backing from Y Combinator and BoxGroup
How it started
Spencer Mateega and Carlos Georgescu founded AfterQuery in San Francisco in January 2025, backed by Y Combinator and BoxGroup. The premise: the internet explains what a cardiologist does, but not how one thinks through an unusual case — so labs need recorded expert reasoning, not scraped text.
What happened
In roughly 14 months the company enrolled nearly 100,000 verified professionals and claims every major frontier lab became a customer. In April 2026 it disclosed a $30M Series A at a $300M valuation led by Altos Ventures, with The Raine Group and angels from Google DeepMind, OpenAI, Anthropic and Meta Superintelligence Labs, and said it had passed a $100M annual revenue run rate.
How it ended up
Still running and scaling as of April 2026: expanding the expert network into more industries and building an enterprise solutions business.
Background
AfterQuery is a San Francisco AI-data startup founded in January 2025 by Spencer Mateega and Carlos Georgescu to sell frontier labs something the internet can't provide: recorded reasoning from verified professionals. Its network of nearly 100,000 doctors, lawyers, financial analysts and software engineers is used to build training datasets and reinforcement-learning environments that teach models how experts actually think through unusual cases.
The bet is that as compute costs fall and algorithms spread through open research, high-quality human-generated training data becomes one of the last real differentiators between models. Within about 14 months AfterQuery signed up its expert network and claims every major frontier lab is now a customer; competitors include Scale AI, Appen and Mercor.
In April 2026 the company announced a $30M Series A at a $300M valuation, led by Altos Ventures with The Raine Group and existing investors Y Combinator and BoxGroup, plus angels from Google DeepMind, OpenAI, Anthropic and Meta Superintelligence Labs. Alongside the round it disclosed that it had passed a $100M annual revenue run rate — reached within 15 months of founding.
What has to be true
- Expert reasoning is scarce and hard to fake: structured recordings of how professionals think can't be derived from public sources the way labeled text can.
- The buyer side moved fast: frontier labs paid for the data at a $100M run rate within 15 months, validating demand early.
- The angel list reads like the AI industry's own bet: insiders from Google DeepMind, OpenAI, Anthropic and Meta put money in.
- The position is still contested: synthetic data pipelines and labs building data operations in-house could erode the moat.
What can be applied
When compute and model recipes commoditize, the scarce input migrates to data — and the fastest wedge is expertise that can't be scraped. Verify demand by watching who pays and how fast.
Aftermath
As of April 2026 AfterQuery continues to operate from San Francisco with a $100M+ annual run rate and roughly 100,000 verified professionals in its network. Its April 2026 Series A values it at $300M — a modest multiple on that run rate — and proceeds are earmarked for growing the expert network, covering more industries, and building an enterprise business. The biggest open risks, flagged in Dealroom's analysis, are labs substituting synthetic data and large customers building data operations in-house.
Sources
- AfterQuery raises $30M at $300M valuation from Altos Ventures as AI labs race for expert training data
- AfterQuery hits $100M run rate as it reveals $30M Series A backed by AI lab insiders
spotted an error? The archive wants to know.
Your turn
You just read one. Describe what you are building, and see who is betting on the same thing.
Free account · 3 free questions · no card