About
Mercor builds an AI-powered expert network and enterprise platform that connects vetted domain specialists with AI labs and businesses for frontier-model training, evaluations, and custom AI agents. Its differentiation is the scale of its human-expertise network, which it organizes into training data, expert evaluations, staffing, and workflow-specific AI systems.
Market
Mercor competes in the human-in-the-loop AI data, model-training and evaluation, and emerging enterprise-agent infrastructure markets. It differentiates by combining an AI-matched domain-expert network and RLHF data with workflow capture, custom agent deployment, quality guardrails, and benchmark infrastructure such as APEX and Archipelago. This positions Mercor as a broader human-expertise-to-agent-performance platform rather than a narrowly focused annotation, evaluation, or recruiting tool.
Mercor primarily serves frontier AI labs and large enterprises, especially Fortune 500 companies, that need expert-generated training data, model evaluation, or AI agents adapted to internal workflows. Likely buyers include AI research and data leaders at labs, plus enterprise AI, operations, recruiting, engineering, and workflow-automation leaders.
At a Glance
Problem
Mercor addresses the bottleneck that advanced AI systems cannot improve reliably without high-quality human judgment, domain expertise, and feedback. Models still struggle with messy real-world workflows, the tools and applications people use, and recognizing what good work looks like in a particular profession or business. For AI labs, this creates a costly, labor-intensive need for expert-generated training, evaluation, and reinforcement data; for enterprises, it makes it difficult to move AI agents from generic demonstrations into economically useful production work. The scale of the pain is substantial: Mercor has reported paying more than $1.5 million per day to people training AI systems.
The central use case is turning scarce professional expertise into usable AI-development data. Lawyers, doctors, engineers, finance professionals, creators, and other specialists evaluate model outputs, complete realistic tasks, and provide the judgments and examples needed to improve models and agents. For enterprises, the same capability helps capture how employees actually work so that agents can be trained against real workflows rather than abstract simulations.
Product / Service
Mercor operates a two-sided marketplace and AI-training infrastructure. It connects professionals with remote projects for leading AI companies, uses an adaptive AI interview to assess expertise at scale, matches approved experts to scoped assignments, and handles payments to contractors across countries and currencies while billing AI labs and enterprise customers. Experts work on tasks such as evaluating, improving, and guiding large language models, with compensation and project terms defined up front.
The company is expanding beyond labor placement into enterprise agent deployment. Mercor says it learns how work gets done inside a company, builds agents equipped with human expertise, and defines what good performance looks like. Its resulting datasets, benchmarks, evaluation criteria, and workflow traces can help labs improve models and help businesses deploy agents with quality controls, while real-world usage feeds new edge cases and performance gaps back into the training process.
Market
Mercor competes across AI data and model-training infrastructure, expert networks, AI-enabled recruiting, model evaluation, and enterprise agent deployment. Its closest named rivals in the AI data race include Scale AI and Surge AI, although Mercor’s positioning increasingly spans both the supply of expert human work and the deployment of agents trained on that expertise. The company’s differentiation is the combination of a large vetted expert network, automated assessment and matching, payments infrastructure, and proprietary data about economically valuable workflows.
Mercor is not pre-revenue. The company reported a $350 million Series C that brought its valuation to $10 billion, and its own materials describe it as a profitable Series C business trusted by six of the seven major technology companies commonly called the Mag 7. Secondary reporting says it reached $2 billion in annualized gross revenue in June 2026, after reaching a $500 million run rate in 17 months. Mercor also says its network includes more than four million vetted experts and 1.9 million referrals, indicating substantial marketplace traction and a potentially reinforcing data and supply-side network effect.
Founders & Leadership
Funding History
General Catalyst
Benchmark
Felicis
Felicis
Recent News
Mercor announced that it will acquire Deeptune, a company building reinforcement-learning environments. The combined business will pair Mercor’s expert network with Deeptune’s software platform to create realistic AI training environments at scale.
Forbes reported that Mercor was in talks to raise $500 million at a $20 billion valuation, although the terms could change and the deal was expected to close later in July. The report also noted that Mercor said its annualized revenue surpassed $2 billion in June.
Mercor introduced its Enterprise AI platform, extending its expertise in frontier-model development to enterprises. The offering is positioned around helping organizations deploy and evaluate AI agents using domain expertise.
Active Roles
94Business Model
Mercor makes money by selling AI training data and enterprise services—including custom AI agents, expert evaluations, and access to vetted domain experts—to AI labs and enterprises. Public evidence does not disclose standardized pricing, so the model appears primarily service- and contract-based rather than a public self-serve subscription.
Products
Customers
Tech Stack
Similar Companies
Competitors
Key Investors
Felicis Ventures, Benchmark, General Catalyst, DST Global, Robinhood, Menlo Ventures, Soma Capital, Link Ventures, 2.12 Angels