AI DATA / OCTOBER 2026

The businesses
behind AI data.

People teach AI how to do useful work. Explore the organizations behind that effort, with sources for their claims.

Explore 39 organizations Find paid work About our proposed lab

Public-source research

39 organizations7 business groups880 source records

Sources and method

THE EVIDENCE

What the numbers show.

How we use evidence

REVENUE

Revenue, with its dates attached.

Revenue covers a stated period. Annualized pace (recent revenue projected across a full year) is an estimate.

REVENUE CHANGE

Two businesses, different revenue changes.

A comparison of each business’s 2024 and 2025 financial years.

WHAT THEY DO

What these organizations sell and build.

We group each organization by its main offering.
Many work across several categories.

Explore a company.

See the full index

Human depth. Machine possibility. A research proposal illustrated with Jung, Einstein and Oppenheimer. Full image text and qualifications follow. — PROPOSED VENTURE
Learning gains and commercial demand remain unproven.
View full image PNG · 2.5 MB · (opens a new tab)

**Our proposed AI Neolab.** A research-first AI lab built around a testable training hypothesis. Original research must demonstrate a measurable training advantage.

57 adult alumni profiles researched
From 170 competition alumni screened (2011–2026); only source-supported adults are counted. Sourcing research, not a signed team. The directory is not published.

$2B Mercor annualized revenue run rate
Run rate (recent revenue projected across a full year). June 2026, published by investor Felicis; gross basis per a Sacra estimate. Not our revenue. See the Mercor record.

Read the image text

ScienceToStartup / A new research thesis

**Human depth. Machine possibility.**

Our alpha thesis

Rare mathematical minds. Original training worlds. Evidence of what transfers.

Math medalists × human feedback × AI research

170
Talent profiles mapped
Sourcing research, not a signed team
$1B+
Mercor annualized revenue run rate
Company-reported, May 2026; not our revenue
$65.5B
AI agents & assistants, 2027 forecast
Gartner, Sep 2026; downstream market, not our TAM

Jung · Einstein · Oppenheimer / conceptual inspirations

Proposed venture. Learning gains and commercial demand remain unproven. sciencetostartup.com

The image reproduces the September 2026 deck. Its figures are superseded by the sourced values above; the $65.5B market forecast could not be traced to a published Gartner release and is not part of our dataset.

Image terms: alpha thesis (an initial idea to test through research); downstream market (the market for products using the proposed work); TAM, total addressable market (the total possible demand for a product).

Paid work listings.

Browse work listings

Company index

See what each organization does. Open its profile for products and evidence.

SearchMain business

More filters

SortDelivery evidence Revenue figures Published work opportunities

Find paid work.

Pay and requirements for public listings.

Rates apply to the listed role.

Search work

Compare

Compare up to three organizations.

THE RESEARCH

How the research is checked.

Each claim links to its evidence. Financial figures retain their dates and the business they cover.

What the labels mean

Delivery evidence (public support for a company’s commercial activity) has several levels. Each rating covers the business described in that record.

Confidence grades

Each claim carries a research grade that says how strong its sources are.

A: Independent or buyer-side evidence, such as a filing or a confirmed customer.
B: Attributed company, investor, or named primary statement.
C: Secondary reporting, an estimate, or indirect evidence.
U: Unverified, undisclosed, or unresolved.

About, contact and corrections

ScienceToStartup is published by Musa Ozdem. It is a public-source research guide to the organizations that produce AI training data. No profiled company has paid for coverage, and none is a customer, investor or affiliate of the publisher.

Disclosure. The publisher is also proposing an AI research lab, described on the overview page. That proposal is separate from the research here, which applies the same rules to every organization.

Corrections. If a figure, date or listing is wrong, email musaeminozdem@gmail.com with the company name and the source. We review within seven days.

Personal data and removal. Company profiles name founders and executives from public company sources only. We do not publish personal contact details. If you are named here and want a record corrected or removed, email the same address.