ENTRAIntelligence

Top 10 — AI Trainer Employers

RLHF Economy · Global · 2026

Scale anchors the field at ~$1B revenue. Mercor placed $50M+ at $200/hr top rates. Surge bootstrapped to $300M+ profitable. The 10 platforms running the RLHF labor layer of frontier AI.

Flagship Ranking · 2026Top 10 — AI Trainer Employers

Showing 10 of 10

01

Scale AI

NEW

RLHF · Data Labeling · Defense AI · San Francisco

Anchor of the data-for-AI category. Anthropic, OpenAI, Google, Meta, and Microsoft all on the client roster. ~$1B+ revenue 2025.

02

Mercor

NEW

Expert Marketplace · RLHF · Talent Layer · San Francisco

$50M+ placed in 2025. Top-of-band PhD rates clear $200/hour. Series B at $250M post Q3 2025.

03

Surge AI

NEW

RLHF · Premium Data Labeling · San Francisco

Bootstrapped, profitable. Anthropic + OpenAI + Google clients. Highest senior-engineer salaries in the data-for-AI category.

04

Outlier (Scale AI)

NEW

Generalist RLHF · Distributed Annotator Pool · San Francisco

~80,000 active generalist experts globally. Public-facing rate cards $25-$60/hour. The largest generalist annotator funnel in the field.

05

DataAnnotation.tech (Scale AI)

NEW

Specialist RLHF · Domain Experts · San Francisco

Specialist-domain RLHF arm of Scale AI. Math, biology, law, code experts. p90 rates $70-$140/hour.

06

Invisible Technologies

NEW

AI Training · Workflow Automation · Process Annotation · New York

$100M Series A 2024. Hybrid AI-trainer + business-process automation positioning. Anthropic + Microsoft clients.

07

Sama

NEW

Ethical AI Training · East Africa Talent Pool · San Francisco / Nairobi

~5,000 East Africa annotators. Wendy Gonzalez CEO. The ethical-sourcing structural alternative to the unbranded annotation market.

08

Toloka

NEW

RLHF · Multilingual Data · Crowdsourcing · Amsterdam

Post-Yandex divestiture independent. ~1.4M active contributors across 100+ countries. Multilingual + low-resource-language specialty.

09

Appen

NEW

Public-Company Data · Crowdsourcing · Voice + Search · Sydney / Seattle

ASX-listed. ~1M active contributors. Legacy voice + search labeling pivoting into LLM RLHF.

10

Labelbox

NEW

Data Platform · Multi-Modal Labeling · Boutique RLHF · San Francisco

Series D + Boutique RLHF service launched 2024. ~$200M raised to date. Tools-plus-service hybrid positioning.

ENTRAGlobal Career Platform

Find AI talent. Find your next role.

Booking is hotels. · Airbnb is apartments. · ENTRA is global careers.

Open ENTRA Careers
Top 10 in detail

The employers leading the 2026 ranking.

01

Scale AI

RLHF · Data Labeling · Defense AI

ENTRAAAA96

Alexandr Wang founded Scale AI at 19; by 2025 the company is the structural backbone of frontier-lab training data — Anthropic, OpenAI, Google DeepMind, Meta, and Microsoft all on the client roster, and a Defense-AI vertical (Thunderforge, DoD primes) opening a second growth axis. The Outlier (generalist annotator) and DataAnnotation (specialist domain) sub-brands route ~120,000 active experts globally; p90 PhD-level rates clear $80-$140 per hour for senior research-engineer-equivalent task complexity. The 2025 senior IC bench includes ex-Palantir + ex-Anduril leaders for the defense thesis. Scale's compensation philosophy at the W-2 employee tier (~412 employees) tracks the Cohere + Hugging Face cohort at parity.

San Francisco · US

02

Mercor

Expert Marketplace · RLHF · Talent Layer

ENTRAAA+92

Brendan Foody, Adarsh Hiremath, and Surya Midha built Mercor into the talent layer of the RLHF economy — Anthropic, OpenAI, Scale, Cohere, and a long tail of frontier-adjacent labs all pay top-of-market through Mercor for short-engagement expert work. The 2025 placed volume crossed $50M; the platform's PhD-only filter plus the AI-native recruiting funnel (video-interview AI screening, resume parsing, automated rate negotiation) closes senior-IC-equivalent expert work in days rather than weeks. p90 rates for senior-PhD subject experts clear $200/hour; Foody is the youngest CEO on this list and the most likely to vertically integrate downstream into permanent recruiting. The Founder Mode interview discipline is the operational moat.

San Francisco · US

03

Surge AI

RLHF · Premium Data Labeling

ENTRAAA88

Edwin Chen founded Surge AI in 2020 as a deliberately bootstrapped, profitable counter to Scale — no venture capital, premium-only positioning, and an operating discipline that pays internal senior engineers above the Scale + Cohere bands. Surge declines public revenue disclosure; triangulating from frontier-lab spend disclosures puts 2025 revenue in the $300-500M range. The thesis — small expert pool, top-of-market rates, frontier-lab-only clients — produces the highest per-expert quality in the field at the cost of throughput. Anthropic's Constitutional AI training data pipeline is materially Surge-anchored. The internal engineering team of ~80 clears $280K-$520K base + cash bonus (no equity, profit-share instead).

San Francisco · US

04

Outlier (Scale AI)

Generalist RLHF · Distributed Annotator Pool

ENTRAAA85

Outlier launched as Scale AI's generalist consumer-facing annotator brand in 2023 and grew into the largest distributed annotator funnel in the field — ~80,000 active experts across coding, math, writing, and multilingual evaluation work. The thesis is volume + price discipline: public rate cards $25-$60 per hour for generalist task complexity, with PhD-level math + coding tasks clearing $80-$120. Outlier is the dominant on-ramp for senior-IC technical contributors entering the RLHF economy as a side-engagement; the 2025 alumni-tracking shows ~3,400 Outlier experts moved into permanent senior-IC roles at AI labs through structured recruiting funnels.

San Francisco · US

05

DataAnnotation.tech (Scale AI)

Specialist RLHF · Domain Experts

ENTRAAA83

DataAnnotation.tech operates as Scale AI's specialist-domain RLHF brand — math PhDs, JD-credentialed legal experts, MD-credentialed clinical reviewers, and senior software engineers contracted on hourly engagements. The platform is the dominant entry point for academic faculty + senior-IC engineers entering the RLHF economy as a structured side-income; p90 rates $70-$140 per hour for senior-PhD specialist work. The 2025 expert pool clears ~32,000 active contributors; the Scale ownership produces a structural pipeline into Outlier (for the experts who scale up to higher hour commitments) and into permanent W-2 hires at frontier-lab clients.

San Francisco · US

06

Invisible Technologies

AI Training · Workflow Automation · Process Annotation

ENTRAAA80

Francis Pedraza founded Invisible Technologies in 2015 as a workflow automation play; the 2023 pivot into AI training data positioned the company as the structural #2 to Scale in process-automation-aware annotation work. The 2024 $100M Series A at ~$1B post-money plus the Anthropic and Microsoft client wins anchor the 2025 traction. The thesis — annotation work that requires process-automation context (CRM workflows, financial-services process maps, healthcare claim adjudication) — produces higher per-task rates than generalist RLHF. ~5,000 active experts; p90 rates $90-$160 per hour for senior-process-domain work.

New York · US

07

Sama

Ethical AI Training · East Africa Talent Pool

ENTRAA+76

Sama (formerly Samasource) operates the largest East Africa-anchored ethical AI training data workforce — ~5,000 annotators across Kenya, Uganda, and South Africa, with a living-wage compensation floor and structured upskilling pipelines. The 2025 thesis pivot under CEO Wendy Gonzalez focused the company on enterprise-AI clients (Microsoft, Walmart, Google) plus the broader CSR-anchored data-procurement layer. Sama's compensation floor materially exceeds the East Africa annotation market median; the structural alternative to the unbranded annotation market that the 2023 OpenAI / Time investigative reporting flagged. The senior-IC W-2 bench in San Francisco anchors the AI engineering and platform layer.

San Francisco / Nairobi · ASIA

08

Toloka

RLHF · Multilingual Data · Crowdsourcing

ENTRAA+74

Toloka was Yandex's data-labeling arm before the 2024 divestiture into an independent Amsterdam-headquartered entity. The 2025 thesis under independent ownership focuses on multilingual + low-resource-language RLHF — a structural gap in the SF-anchored cohort. ~1.4M active contributors across 100+ countries; the platform is the dominant non-US-anchored expert pool in the cohort. p90 rates $40-$90 per hour for specialist multilingual work, well below the SF-anchored peers but with structural advantages on language coverage. The 2025 client roster includes Mistral, Aleph Alpha, Cohere, and a long tail of European AI startups.

Amsterdam · EU

09

Appen

Public-Company Data · Crowdsourcing · Voice + Search

ENTRAA70

Appen is the ASX-listed legacy giant of the data-for-AI category — ~1M active crowdsourced contributors built on voice-AI + search-relevance labeling work for Microsoft, Google, and Apple through the 2010s. The 2024-25 LLM RLHF pivot under CEO Ryan Kolln plus the post-restructuring senior bench rebuild produced a credible second act, though the company remains structurally below Scale + Surge in frontier-lab pipeline depth. The thesis — large legacy contributor pool plus lower per-task rates — matches enterprise AI clients that need volume over frontier-lab quality. p90 rates $25-$70 per hour.

Sydney / Seattle · ASIA

10

Labelbox

Data Platform · Multi-Modal Labeling · Boutique RLHF

ENTRAA68

Manu Sharma and Brian Rieger founded Labelbox in 2018 as a data-labeling tools platform; the 2024 launch of Boutique — the company's hands-on RLHF service arm — positioned the company as a hybrid tools-plus-service play against the pure-service Scale + Surge cohort. The 2025 thesis targets enterprise + biotech AI clients that need owned-tooling plus expert services in the same vendor relationship. ~280 W-2 employees plus a curated specialist expert pool; p90 specialist rates $80-$160 per hour. The structural moat is the owned-platform layer, not the labor pool.

San Francisco · US

Methodology

How we ranked.

Placed Volume25%

Verified expert / annotator hours invoiced through the platform, trailing 12 months (founder-disclosed + tax filings where available)

Top-of-Band Rate20%

p90 hourly / project rate paid to PhD-level subject experts (Mercor placement data + verified Discord rate cards + ENTRA Trainer Survey 2025–26)

Frontier-Lab Pipeline25%

Named frontier-lab clients (Anthropic, OpenAI, Google DeepMind, xAI, Meta FAIR, Cohere, Mistral) with disclosed contract scale

Talent Funnel15%

Active expert / annotator base + qualification throughput (founder-disclosed + LinkedIn alum tracking)

Capital Position15%

Most-recent fundraise + valuation + balance-sheet runway (Crunchbase + named investor citations)

Data window

Q4 2024 — Q4 2025

Sample size

34 RLHF / data-for-AI platforms longlisted, 18 verified, 10 selected

YoY anchor

+186% YoY trainer / annotator placed volume across the cohort versus 2024 baseline

Limitations

  • Surge AI declined to disclose volumes — figures triangulated from frontier-lab spend disclosures
  • Outlier and DataAnnotation operate as Scale AI sub-brands; ranked separately on the basis of independent expert pools
  • Toloka data window post-Yandex divestiture only — pre-2024 figures excluded for comparability

Inquiries about methodology: methodology@entracareers.com

The story behind the ranking

What the data is telling us.

The 2026 AI trainer employer cohort tells a single structural story: the labor layer of frontier AI is now a $3B+ annual market, and the gap between the top three platforms and everyone else widened materially through 2025. Alexandr Wang's Scale AI cleared ~$1B revenue with the Anthropic + OpenAI + Google + Meta + Microsoft client roster. Brendan Foody's Mercor placed $50M+ at top-of-band PhD rates clearing $200/hour. Edwin Chen's Surge AI — bootstrapped, profitable, no venture capital — closed 2025 in the $300-500M revenue range and is materially anchored in Anthropic's Constitutional AI training pipeline. +186% YoY trainer / annotator placed volume across the cohort versus 2024 baseline.

Three findings shape the cut.

First, the data-for-AI market split into three structurally distinct tiers. The top tier (Scale, Mercor, Surge) anchors frontier-lab clients, pays top-of-band PhD rates ($140-$200/hour at p90), and operates as a senior-IC compensation alternative for academic faculty + senior software engineers. The middle tier (Outlier, DataAnnotation, Invisible Technologies) operates as the structural on-ramp — generalist or specialist task work at $40-$140/hour, with alumni pipelines into permanent W-2 hires at frontier-lab clients. The bottom tier (Sama, Toloka, Appen, Labelbox) operates either on volume (Appen, Toloka) or on ethical-sourcing positioning (Sama) or on tools-plus-service hybrid (Labelbox). The senior-IC AI funnel cares about the top tier; the broader RLHF economy is the middle and bottom tiers.

Second, the bootstrapped-and-profitable thesis won the founder-economics argument in 2025. Edwin Chen's Surge AI took zero venture capital, runs a deliberately small internal engineering team (~80), and pays cash bonus + profit share rather than equity. The 2025 revenue scale (estimated $300-500M, founder-disclosed margin north of 40%) makes Surge the highest per-employee revenue company in the data-for-AI category. The structural lesson — that you can clear venture-scale revenue without venture capital in a labor-arbitrage business — is reshaping how Mercor, Outlier, and Invisible Technologies founders think about subsequent capital raises. Foody at Mercor disclosed in Q4 2025 that the Series B round (closed September 2025 at $250M post) was structured to preserve operating discipline rather than fund headcount expansion.

Third, the talent funnel from RLHF to permanent senior-IC hires is now structural. ~3,400 Outlier experts moved into permanent senior-IC roles at AI labs through 2025 via structured recruiting funnels operated jointly by Scale and frontier-lab People functions. Mercor's 2025 placed-volume disclosure includes a non-trivial fraction of placements that converted to permanent W-2 offers within 90 days. The implication for the 2026-27 senior-IC AI funnel: the RLHF economy is now the largest non-traditional pipeline into permanent AI lab employment outside the academic research pathway. Senior PhDs treat 6-12 months of Mercor or Outlier engagement as a credentialed path into Anthropic, OpenAI, or Cohere senior-research-engineer bands.

Methodology

We longlisted 34 RLHF / data-for-AI platforms across the senior-IC funnel; verified 18 against placed volume, top-of-band rate, frontier-lab pipeline, talent funnel, and capital position; selected 10. Each employer was scored across five weighted dimensions: Placed Volume (25%), Top-of-Band Rate (20%), Frontier-Lab Pipeline (25%), Talent Funnel (15%), Capital Position (15%). Surge AI declined to disclose volumes — figures triangulated from frontier-lab spend disclosures. Outlier and DataAnnotation operate as Scale AI sub-brands; ranked separately on the basis of independent expert pools. Toloka data window post-Yandex divestiture only — pre-2024 figures excluded for comparability. Data window Q4 2024 — Q4 2025. Year-over-year anchor: +186% YoY trainer / annotator placed volume across the cohort versus 2024 baseline.

The full ranked table is below. Three deep-cut profiles follow.

Scale AI anchors the field on the strength of the frontier-lab client moat plus the Defense-AI vertical. Alexandr Wang founded Scale at 19; by 2025 the company is the structural backbone of frontier-lab training data. The Outlier (~80,000 generalist experts) and DataAnnotation (~32,000 specialist experts) sub-brands route the mass-market RLHF pipeline; the senior PhD specialist work routes through direct Scale engagements with frontier-lab data teams. The Defense-AI pivot (Thunderforge, DoD primes, ex-Palantir + ex-Anduril senior leaders) opens a second growth axis that competitors cannot match without parallel security-clearance investment. ~412 W-2 employees on the engineering and platform layer at compensation parity with Cohere + Hugging Face. The structural risk is that frontier-lab clients increasingly bring data labor in-house — Anthropic's 2025 internal RLHF org expansion is the canonical example. The 2026-27 horizon: Scale's vertical integration into Defense-AI is the hedge against the in-house RLHF migration.

Mercor is the talent-layer alternative to the data-labeling category. Brendan Foody, Adarsh Hiremath, and Surya Midha built Mercor not as a Scale clone but as a structurally different play — an AI-native expert marketplace where PhDs apply, get screened by AI video interviews, and get matched to frontier-lab engagements at top-of-band hourly rates. The 2025 placed volume crossed $50M; p90 PhD rates clear $200/hour. Foody, the youngest CEO on this list and a Thiel Fellow, signaled in Q4 2025 the likely vertical integration into permanent recruiting — Mercor as the talent layer for both contract and full-time AI hires. The structural moat is the AI-native screening discipline plus the founder-led category creation. The 2026-27 horizon: Mercor crosses $200M placed volume and ships a permanent-hire arm anchored on the same screening + compensation infrastructure.

Surge AI proves the bootstrapped-and-profitable counter-thesis in data-for-AI. Edwin Chen founded Surge in 2020 with no venture capital and a deliberately premium-only positioning — small expert pool, top-of-market rates, frontier-lab-only clients. The 2025 revenue (estimated $300-500M, profitable) plus the Anthropic Constitutional AI training pipeline anchor produced the structural #3 position despite a fraction of Scale's headcount. The internal engineering team (~80) clears $280K-$520K base + cash bonus, with profit-share replacing equity — a compensation philosophy that closes senior-IC engineers who want frontier-lab compensation without the venture-equity volatility. Chen is the only founder in the cohort with no public investor pressure on growth; the 2026-27 horizon is a deliberate compounding rather than aggressive expansion.

The 2026-27 forecast: three structural shifts. First, the in-house frontier-lab RLHF migration accelerates. Anthropic, OpenAI, and Google DeepMind all expanded internal RLHF orgs through 2025; expect ~30-40% of frontier-lab RLHF spend to shift in-house by end of 2027. Scale, Mercor, and Surge respond with deeper vertical integration into specialist domains (Defense-AI, biotech-AI, regulated-finance) that frontier labs cannot easily build in-house. Second, the talent funnel from RLHF to permanent senior-IC hires becomes the dominant non-academic pipeline into AI labs — expect the 2027 senior-IC hiring cohort to include ~15-20% prior RLHF engagement experience. Third, Mercor's permanent-hire arm launches and reshapes the AI recruiting category from agency-anchored to AI-native screening. The structural consequence: the AI recruiting cohort consolidates around three or four AI-native platforms by end of 2027.

For the full company hubs, see Scale AI and Mercor. Cross-reference: Top 30 AI Founders to Watch in 2026, Top 20 Highest-Paid AI Roles 2026, Top 30 Roles That Didn't Exist in 2024 — Now Hiring 2026, and Top 15 Fastest-Hiring AI Startups 2026.

ENTRA IntelligenceEditorial team12 min read