Mercor

TL;DR. Mercor is a San Francisco AI talent and data-services company founded in 2023 that supplies vetted domain experts to frontier AI labs for RLHF, model evaluation, and agent training. It serves the top five AI labs and six of the Magnificent Seven, reaching a $10 billion valuation in October 2025.

Overview

Mercor (Mercor.io Corporation) is a San Francisco-based AI company that supplies vetted domain experts to frontier AI labs for model training, evaluation, and reinforcement learning. Its mission, as stated on its website, is "organizing human intelligence to power the AI economy."

Founding and leadership. Mercor was founded in 2023 by Brendan Foody (CEO), Adarsh Hiremath (CTO), and Surya Midha (board chairman), three friends from Bellarmine College Preparatory in San Jose who competed together on the school's Speech and Debate team. All three dropped out of college and were awarded Thiel Fellowships. In May 2025, the company hired former Uber chief product officer Sundeep Jain as its first president. All three founders appeared on the Forbes 30 Under 30 list for 2025 and became the world's youngest self-made billionaires at age 22 following the company's Series C.

Pivot and market positioning. Mercor originally connected freelance software engineers in India with U.S. companies using AI-driven resume screening and AI-avatar interviews. After conversations with OpenAI and xAI, the founders pivoted toward expert-driven AI evaluations and data labeling, matching specialized professionals (scientists, doctors, lawyers, bankers) with AI labs for RLHF and agent training. The company benefited significantly when OpenAI and other labs reduced reliance on Scale AI following Meta's $14 billion investment in Scale and hiring of its CEO. Mercor claims to supply expert contractors to five top AI labs including OpenAI and Anthropic.

Funding. General Catalyst led the $3.6 million seed round in 2023. Benchmark led the $32 million Series A in 2024 at a $250 million valuation. Felicis led the $100 million Series B in February 2025 at a $2 billion valuation, and led the $350 million Series C in October 2025 at a $10 billion valuation (5x the Series B). Benchmark, General Catalyst, DST Global, Menlo Ventures, and Robinhood Ventures have participated across rounds. Notable individual backers include Peter Thiel, Jack Dorsey, and Adam D'Angelo.

Scale and operations. Headquartered at 181 Fremont, San Francisco, Mercor has approximately 300 employees. The company reports a network of over 4 million vetted experts, pays more than $2 million per day to contractors, and told investors it was on track to reach $500 million in ARR. In January 2026, Mercor released its first APEX-Agents benchmarking report evaluating how leading AI models perform business tasks in consulting, investment banking, and law.

Key links. Website: mercor.com; Careers: mercor.com/careers; Blog: mercor.com/blog.

Products & Features

Mercor operates three interconnected product lines: an AI-powered talent marketplace that sources and matches domain experts to AI training work, a frontier AI research division producing benchmarks and evaluation datasets, and an enterprise platform for building custom AI agents. The company positions itself as "organizing human intelligence to power the AI economy," serving the top 5 AI labs and 6 of the Mag 7 companies.

Talent Marketplace and AI Interview Candidates upload a resume and complete a roughly 20-minute AI-conducted video interview. The AI interviewer asks role-specific questions, generates a transcript, and evaluates performance to build a scored profile. Once vetted, candidates enter a talent pool where Mercor matches them to client projects. The platform supports domain experts across medicine, law, finance, consulting, engineering, and more. Listed expert rates range from $60/hr to $180/hr depending on specialization (e.g., hematology/oncology at $130-$180/hr, management consultants at $100/hr). Candidates can retake interviews up to three times. The web platform is accessible via Chrome, Edge, and Safari (not mobile or Firefox).

AI Data Services for Labs Mercor supplies frontier AI labs with post-training data services including RLHF, supervised fine-tuning (SFT), model evaluation, rubric development, and reinforcement learning environment construction. The company operates on a "cost-plus" hourly pricing model where Mercor and the client jointly set expert compensation and Mercor takes a fixed percentage margin. Mercor recommends hourly pay over per-task pay to optimize for data quality and iteration speed.

APEX Benchmark Family Mercor Research publishes a suite of public AI benchmarks. APEX evaluates models on economically valuable tasks across investment banking, management consulting, big law, and primary care medicine. APEX-SWE, built with Cognition, tests software engineering tasks. APEX-Agents measures long-horizon, cross-application agentic tasks in professional services. ACE (AI Consumer Index) tests everyday consumer tasks. Benchmarks feature public leaderboards, are graded by domain experts and judge LMs, and datasets are published on Hugging Face. An accompanying arXiv paper describes APEX-v1.0 with results for 23 models from 13 providers.

Enterprise and Data Partnerships Mercor Enterprise offers four services: Agent Diagnostics (workflow analysis and roadmap), Agent Deployment (custom agent building), Agent Benchmarking (independent product testing), and Data Monetization (licensing anonymized enterprise workflow data to AI labs). The data partnerships product connects to 34+ enterprise tools via OAuth, extracts workflow data, and applies proprietary anonymization that detects and masks 60+ types of sensitive identifiers. The platform holds SOC 2 Type II certification and uses AWS KMS encryption. Enterprise deployment timelines run 4-6 weeks.

Mercor reports a $10B valuation, $2B+ revenue run rate, 100k+ contractors hired, and 400+ full-time employees.

Security & Compliance

Mercor experienced a major data breach in March 2026 via a supply chain attack on the open-source LiteLLM library, resulting in the theft of approximately 4 TB of data. The breach exposed the personal information of over 40,000 contractors, including Social Security numbers, video interview recordings, identity verification documents, platform source code, internal Slack communications, and API keys. The extortion group Lapsus$ claimed responsibility and posted samples of stolen data on its leak site.

The attack originated when threat actor group TeamPCP compromised LiteLLM's CI/CD pipeline (tracked as CVE-2026-33634) and published malicious versions 1.82.7 and 1.82.8 to PyPI on March 27, 2026. The packages were live for roughly 40 minutes to 3 hours. When Mercor's systems ingested the poisoned packages, credential-harvesting malware executed and stole API keys, cloud credentials, SSH keys, and database passwords, enabling lateral movement through Mercor's infrastructure. Mercor confirmed the incident on March 31, 2026, telling TechCrunch it was "one of thousands of companies" affected.

The breach collided with a separate scandal involving Delve Technologies, a GRC startup that had issued LiteLLM's SOC 2 and ISO 27001 certifications. A whistleblower alleged Delve fabricated compliance reports with 99.8% identical text across all clients, routing audits through certification mills. It is unclear whether Mercor itself held SOC 2 or ISO 27001 certifications, or whether it relied on LiteLLM's Delve-issued certifications as part of its vendor trust model.

At least five class action lawsuits were filed within one week of the breach. Plaintiffs allege Mercor failed to implement multifactor authentication, did not encrypt sensitive data at rest or in transit, lacked proper access controls, and failed to monitor for suspicious activity. Meta indefinitely paused all work with Mercor following the breach. Mercor's official Trust Center (trust.mercor.com) and its privacy policy state that the company implements "appropriate technical and organizational measures" and complies with GDPR, but the Trust Center page returned no substantive content as of this research.

Mercor's privacy policy, last updated August 27, 2025, details collection of interview recordings, transcriptions, profile photos, and salary expectations. It references international data transfer safeguards including standard contractual clauses and adequacy decisions. Contractors sign a Confidential Information and Inventions Assignment Agreement (CIIAA) and Terms of Work, accessible through the platform's dashboard. The legal entity is Mercor.io Corporation, based in San Francisco.

User Feedback & Adoption

Mercor has rapidly scaled to over 30,000 active contractors across 45+ countries as of late 2025, collectively paid more than $1.5 million per day for AI training work. The platform holds a 4.0 rating on Glassdoor (116 reviews) and a 4.0 on Trustpilot (218 reviews), reflecting a polarized user base: contractors who secure steady projects praise the pay and flexibility, while those who experience gaps in work availability are disproportionately negative.

Common praise themes. Reliable, weekly Stripe payments are the most consistently cited positive, with reviewers across Trustpilot, Glassdoor, and Indeed confirming on-time payouts. Competitive hourly rates (advertised range of $50 to $200 per hour, with a platform-stated average around $85 to $95 per hour) rank as the top attraction, earning a 4.1/5 compensation sub-rating on Glassdoor. Fully remote, flexible scheduling and intellectually stimulating work training frontier models for labs like OpenAI and Anthropic are also frequently highlighted.

Common complaint themes. "Feast or famine" work availability is the dominant grievance: projects are paused or canceled abruptly without notice, and accepted offers are sometimes "paused indefinitely" with no follow-up. Contractors report poor-to-nonexistent communication from management, feeling disposable, and over-onboarding (too many contractors accepted for limited task pools). The mandatory time-tracking software (Insightful/Workpuls), which requires camera, microphone, and browser-history access, is widely described as a privacy intrusion. The one-way AI interview process draws criticism for being impersonal, with some applicants suspecting it functions as a data-harvesting exercise rather than genuine hiring. As 1099 contractors, workers receive no benefits and bear self-employment taxes.

Notable adoption metrics. Mercor's revenue run rate surged from $1 million to $500 million in approximately 17 months, driven by its 30,000+ contractor base. India is the largest talent source, followed by the United States. The company raised $350 million in a Series C round in October 2025 at a $10 billion valuation, with investors including Felicis, Benchmark, General Catalyst, and Robinhood Ventures. Active legal challenges, including a Scale AI trade secrets lawsuit and a worker misclassification class action, add uncertainty for contractors weighing long-term commitment.

Monetization & Business Model

Mercor operates a technology-enabled talent marketplace, charging AI labs and enterprises a percentage-based take rate on contractor placements for AI model training work. The company typically charges around 30% of a contractor's pay rate as a finder's fee and matching fee, though the take rate varies by client and engagement complexity. Contractors receive roughly 60 to 70% of gross top-line revenue, with experts earning over $85 per hour on average. Mercor's revenue figure reflects total customer spend before contractor payouts (gross revenue), a practice CEO Brendan Foody noted is standard among competitors like Scale AI and Surge AI. As of late October 2025, Mercor was paying over $1.5 million per day to its contractor network of more than 30,000 experts.

The company's revenue trajectory has been extraordinary. ARR reached $75 million in February 2025, $100 million in March 2025 (posted by Foody on X), and was approaching $450 million in annualized run-rate revenue by September 2025. Mercor told investors it was on track to hit $500 million ARR faster than Anysphere (Cursor), which reached that milestone roughly a year after product launch. Sacra estimates Mercor's annualized gross revenue hit $2 billion by June 2026. Unlike many high-growth AI startups, Mercor was profitable early, generating $6 million in profit in H1 2025.

Funding totals approximately $486 million across four rounds. The September 2024 Series A raised $30 million at a $250 million valuation. The February 2025 Series B raised $100 million at a $2 billion valuation, led by Felicis with Benchmark, General Catalyst, and DST Global participating. The October 2025 Series C raised $350 million at a $10 billion valuation (5x the Series B), led by Felicis with Benchmark, General Catalyst, and new investor Robinhood Ventures. Notable early investors include Peter Thiel, Jack Dorsey, Adam D'Angelo, and Larry Summers. Mercor's growth accelerated significantly after Meta's $14.3 billion investment in Scale AI prompted OpenAI and Google DeepMind to seek alternative data-labeling partners.

Leadership & Team

Mercor was founded in 2023 by three Bay Area high school friends who dropped out of college to build the company. All three are Thiel Fellows and appeared on the Forbes 30 Under 30 list for 2025.

The three co-founders met at Bellarmine College Preparatory in San Jose, where they competed together on the speech and debate team. All are children of software engineers and grew up in the Bay Area. In October 2025, Mercor's $350 million Series C at a $10 billion valuation made all three (then 22 years old) the world's youngest self-made billionaires, each holding roughly a 22% stake, surpassing Mark Zuckerberg who debuted as a billionaire at age 23.

Jain was hired in May 2025 as Mercor's first president, bringing deep operational experience from Uber and Google to complement the founders' youth. The company has approximately 300 employees as of 2025. The founders describe an intense work culture, with Foody reporting he leaves the office around 10:30 p.m. six days a week. Notable investors include Benchmark, Peter Thiel, General Catalyst, Felicis Ventures, Robinhood Ventures, and Larry Summers.

Target Audience & Use Cases

Mercor serves two distinct audiences: frontier AI labs that need domain expertise to train and evaluate models, and Fortune 500/2000 enterprises deploying custom AI agents in production. The platform operates a two-sided marketplace connecting these demand-side customers with a supply-side network of over 100,000 vetted domain experts spanning 300+ professional fields.

Demand side: AI labs. Mercor's primary customers are the world's leading AI labs, including OpenAI, Anthropic, Google DeepMind, and Meta, along with six of the seven "Magnificent Seven" technology companies. These labs use Mercor to source expert labor for post-training tasks: RLHF data collection, rubric creation, data labeling, and AI agent evaluation. CEO Brendan Foody has stated that most of Mercor's revenue comes from a small number of AI lab clients. The labs pay premium rates, with Mercor charging marked-up hourly billing on contractor engagements.

Demand side: Enterprises. Through Mercor Enterprise, the company extends the same expert network and evaluation infrastructure to Fortune 500 companies deploying AI agents. Enterprise offerings include agent diagnostics (workflow mapping to identify high-value use cases), agent deployment, agent benchmarking, and enterprise data monetization. Target verticals include law, finance, medicine, accounting, HR, software engineering, and market research.

Supply side: Domain experts. The expert network includes consultants, lawyers, physicians, scientists, engineers, and finance professionals, vetted through AI-driven interviews. Experts are paid an average of over $95 per hour, with some specialists earning up to $200 per hour. Mercor distributes over $4 million daily to contractors. The platform recruits former employees of investment banks, consulting firms, and law firms whose domain knowledge helps labs train models for complex professional workflows.

Company size and geography. On the demand side, Mercor targets large enterprises and well-funded AI labs rather than SMBs. The company is U.S.-headquartered but recruits experts globally, having originated as a marketplace connecting programmers in India with U.S. companies. Mercor was recognized in the Forbes AI 50 (2025), Forbes Cloud 100 (2025), and Bloomberg's 24 AI Startups to Watch (2026).

Tags & Categories

Primary Category: AI Other Categories: AI Other, AI HR, AI Analytics Tags: talent marketplace, ai training data, rlhf, data labeling, expert annotation, ai model evaluation, human-in-the-loop, global talent sourcing, ai labs, remote contracting, domain experts, benchmark evaluation

Mercor is an AI-powered talent marketplace that connects organizations (primarily frontier AI labs) with vetted domain experts for AI training, evaluation, and reinforcement learning from human feedback (RLHF) work, while also operating a hiring platform for general professional roles. Its core value proposition spans three interrelated areas: expert sourcing for frontier data generation, AI-powered recruitment matching, and benchmark-based AI evaluation research.

The platform serves the top 5 AI labs and 6 of the 7 "Mag 7" technology companies, positioning itself as neutral data infrastructure after major labs sought alternatives to Scale AI following Meta's $14.3B investment in that company. Mercor's expert marketplace covers diverse domains including medicine, law, finance, energy, telecom, and software engineering, with role listings ranging from $60/hr to $180/hr for specialized experts.

Key competitors include Surge AI (reportedly the largest by revenue, crossing $1B run rate in mid-2025), Scale AI, Turing, Invisible Technologies, and Micro1. Mercor differentiates itself through AI-powered candidate vetting (including AI avatar interviews), semantic matching, and a focus on sourcing highly credentialed domain experts rather than traditional crowd-based data labeling.

Beyond talent matching, Mercor develops proprietary AI benchmarks (APEX, APEX-Agents, APEX-SWE, ACE) that measure whether frontier AI models can perform economically valuable professional tasks. The company also publishes peer-reviewed research and evaluation leaderboards, adding a research dimension to its marketplace model.

Impact & Recommendations

Mercor has emerged as the fastest-growing company in the AI data infrastructure market, scaling from $1 million to $500 million in annualized revenue run rate in just 17 months and reaching a $10 billion valuation in October 2025. The company occupies a unique position at the intersection of AI talent supply and frontier model training, paying over $1.5 million daily to a network of 30,000+ expert contractors who provide RLHF data and human evaluation for clients including OpenAI, Google DeepMind, Meta, Microsoft, Amazon, and Nvidia.

Competitive Landscape & Moat. Mercor's closest competitors are Scale AI, Surge AI, and Turing. Scale AI remains the largest player but suffered a significant customer exodus after Meta invested $14.3 billion for a 49% stake and hired CEO Alexandr Wang in June 2025, creating a neutrality concern for rival AI labs. Mercor was a primary beneficiary of that disruption. Surge AI reportedly crossed $1 billion in revenue in 2024 without VC backing and is exploring a $25 billion valuation. Micro1 is Mercor's closest analogue in focusing on high-skill expert matching. Mercor's moat lies in its AI-driven interview and matching pipeline, which enables rapid sourcing of domain experts (doctors, lawyers, scientists) at scale, a capability that traditional staffing firms and lower-skill data labeling shops cannot easily replicate.

Growth Signals. Revenue grew fivefold between March and September 2025. The $350 million Series C (led by Felicis, with Benchmark, General Catalyst, and Robinhood Ventures) quintupled the valuation from $2 billion in just eight months. Sacra estimates Mercor's annualized revenue reached $1.5 billion by May 2026. The company's APEX research index, measuring AI capability across professional roles, signals an ambition to become the definitive benchmark for AI-driven economic work.

Risk Factors. Scale AI filed a federal lawsuit in September 2025 alleging trade secret theft by a former employee who joined Mercor. Contractor relations present a material reputational risk: Forbes reported that Mercor abruptly cancelled a 5,000-person Meta project ("Musen") in November 2025 and rehired workers at $16/hour, down from $21, representing a 24% pay cut. Reddit forums and press coverage have surfaced worker complaints about surveillance, unpredictable hours, and abrupt offboarding. Worker lawsuits over data collection practices have also been reported. These tensions could hinder talent retention and attract regulatory scrutiny.

ICP Fit & Outlook. Mercor is best suited for frontier AI labs and large enterprises that need high-skill human feedback at scale for RLHF, model evaluation, and agent training. For the Chiri Atlas audience, Mercor is relevant as a critical infrastructure layer in the AI stack rather than a general-purpose tool. The outlook is strongly positive on growth trajectory but tempered by litigation, contractor satisfaction risks, and the cyclicality of AI training demand. If AI labs insource human evaluation or reduce reliance on third-party data vendors, Mercor's revenue concentration among a handful of clients becomes a vulnerability. Recommendation: high relevance for organizations building or fine-tuning frontier models; monitor contractor relations and legal outcomes as key risk indicators.

Chiri Analysis

Chiri Score: 72/100

Verdict

Best for:

• Frontier AI labs needing high-skill human feedback at scale for RLHF, SFT, and model evaluation

• Fortune 500 enterprises deploying custom AI agents and monetizing anonymized workflow data

• Credentialed domain experts (doctors, lawyers, bankers, consultants) seeking remote, hourly AI-training work at $60-$180/hr

• Organizations building or fine-tuning frontier models that want a neutral alternative to Scale AI

Not for:

• Organizations requiring a vendor with a proven, unblemished security and compliance track record after the 2026 breach

• SMBs or teams needing low-cost, general-purpose data labeling rather than expert-driven work

• Contractors seeking stable, guaranteed hours or W-2 benefits given 'feast or famine' availability and 1099 status

• Buyers uncomfortable with revenue concentration among a handful of AI-lab clients

Head-to-head

Frequently Asked Questions

What does Mercor do?

Mercor supplies vetted domain experts to frontier AI labs and enterprises for model training, evaluation, and reinforcement learning from human feedback (RLHF). It runs an AI-interview talent marketplace, lab data services, the APEX benchmark family, and an enterprise AI-agent platform.

Who are Mercor's customers?

Mercor serves the top 5 AI labs and 6 of the 7 Magnificent Seven companies, including OpenAI, Anthropic, Google DeepMind, Meta, Microsoft, Amazon, and Nvidia.

Is Mercor SOC 2 compliant?

Mercor's enterprise data product claims SOC 2 Type II certification, but its status is unclear following the March 2026 breach, which was linked to the LiteLLM library whose SOC 2 and ISO 27001 certifications were issued by Delve Technologies, a firm a whistleblower accused of fabricating compliance reports.

Did Mercor have a data breach?

Yes. In March 2026 Mercor suffered a supply-chain attack via the open-source LiteLLM library, exposing approximately 4 TB of data including personal information of over 40,000 contractors. Lapsus$ claimed responsibility, at least five class actions followed, and Meta paused all work with Mercor.

How much does Mercor cost and how does it charge?

Mercor uses a cost-plus hourly model, typically taking around 30% of a contractor's pay rate as finder's and matching fees, with the rate varying by client. Contractors keep roughly 60-70% of gross revenue, and expert rates range from $60 to $180 per hour.

Who founded Mercor and how big is it?

Mercor was founded in 2023 by Brendan Foody (CEO), Adarsh Hiremath (CTO), and Surya Midha (board chairman), three Bellarmine College Preparatory friends and Thiel Fellows. It reached a $10 billion valuation in October 2025 and pays over $1.5 million per day to 30,000+ contractors.

Who are Mercor's competitors?

Mercor's main competitors are Scale AI, Surge AI, Turing, Invisible Technologies, and Micro1. Mercor gained share after Meta's $14.3 billion Scale AI investment prompted labs to seek neutral alternatives.

Is Mercor good for enterprise use?

Mercor is well suited for frontier AI labs and Fortune 500 enterprises deploying AI agents, offering agent diagnostics, deployment, benchmarking, and data monetization. However, the 2026 breach and related lawsuits are material security considerations for enterprise buyers.

---

Reviewed by Chiri Atlas Research Desk (AI Tooling Analyst) on 2026-07-08.