micro1 supplies AI labs and enterprises with expert-made training data, RL environments and AI agent evaluation, and recruits the domain experts behind that work through its Zara AI interviewer.
Used this tool? Rate it
Used this tool? Rate it
micro1 is a data company for AI developers that describes itself as building the infrastructure for advancing intelligence through expert human data, real-world training environments and contextual evaluations. In practice it sells AI training data made by vetted experts, RL environments and benchmarks for frontier models, and AI agent evaluation.
Most of it is not a self-serve app. Buyers start with a contact form that micro1 reserves for companies and clients, while anyone who wants paid work is sent to a separate opportunities page instead. The people doing that work are recruited through Zara, micro1's AI recruiter agent, which sources and vets domain experts at high velocity to supply the human side of new expert data.
A technology news report describes micro1 as having begun as an AI recruiting startup. In its Series A announcement dated September 12, 2025, micro1 said it raised $35M at a $500M valuation. In August 2026 the same news outlet, citing a person familiar with the company rather than micro1 itself, reported that its gross annual run rate grew from $100 million to $500 million over eight months.
micro1's scale figures are self-reported: it says tens of thousands of experts actively contribute to AI training projects through it, and its government page claims 130,000+ deeply vetted candidates across 100+ domains and 60+ languages.
Together with Zara, these form a human-in-the-loop pipeline, and micro1 says its data platform measures velocity, error rates, cost per task and quality in real time. Not every dataset is hand-made, though: a news report says micro1 is increasingly generating synthetic data without human involvement, such as automated descriptions of video content.
Cortex is the AI agent evaluation product, described as a way to evaluate, improve and monitor AI agents with expert human judgment derived from real workflows. It frames human expert review as going beyond automated evals and LLM-as-judge scoring. In micro1's words, Cortex evaluates real workflows, diagnoses failures, creates targeted training data and monitors performance, in four stages:
micro1 lists open-ended technical questions tied to the chosen role, scenario-based questions, a coding challenge for technical roles and a human data exercise for annotator roles. Candidates get a real-time result on whether they meet micro1's global certification criteria, with the option to request feedback if they are not certified. micro1 also says no two candidates receive the same questions.
Interview Prep is the part anyone can try without a sales conversation: you pick a position, take an AI-powered practice interview and get feedback on your performance, and micro1 calls it 100% free. Role cards on the page are labeled as 18-minute interviews.
A separate program pays companies for their internal know-how: micro1 helps companies monetize operational data such as SOPs, knowledge bases and CRM data through data licensing opportunities.
micro1's research page lists Realm benchmarks for legal, pathology-report and tax reasoning. On the government side, micro1 says it achieved Awardable status through the Chief Digital and Artificial Intelligence Office's Tradewinds Solutions Marketplace.
The expert side works like a board of remote AI training jobs: micro1 promises fully remote, flexible work and lists the path in four steps: apply for opportunities, take the AI interview, get onboarded and start earning. After applying, you can check the status of your application directly in your talent dashboard.
micro1's advice is to choose one core skill or job type to be evaluated on and to use the Interview Prep tool to simulate interviews before taking the real one.
According to the recording pages, contributors who submit recordings and annotations get paid on a biweekly basis.
For data, evaluation or hiring work, the contact form asks for contact details, a description of the project and where you heard about micro1. For a data partnership, micro1 first reviews the submission for fit; if there's a potential fit, it requests more information and schedules a discovery call to discuss data and privacy requirements before any agreement.
The official pages checked for this listing do not publish prices for Realm, Cortex, robotics data or Zara hiring, and the old pricing address now lands on the homepage. Buyers go through the company contact form, and Interview Prep is free. Money flows the other way for contributors and data partners:
Experts see the pay rate before accepting a project, and the support program says they are paid on a fixed, predictable schedule twice a month, while the recording pages say biweekly. Company referral rewards are paid once the referred company reaches specific revenue milestones with micro1, and partnership compensation is based on the quality, uniqueness and value of the data provided.
As captured on September 25, 2026, micro1's profile on a third-party review platform showed 10,465 reviews and an overall score of 4.7 in the Recruitment Service category. The profile is claimed by the company and marked as a paid subscription on that platform.
The platform's AI-generated summary of 9,919 reviews is strongly positive, but it notes that some reviewers felt rushed during the assessment or preferred to talk to real people, and a few reported technical hiccups, interview interruptions or delays in automated responses. micro1's own live candidate experience tracker showed 4.34/5 from 465,974 reviews on the same date; that figure is self-reported.
Only partly. Interview Prep is free, but Realm, Cortex, robotics data and data partnerships all start with a contact form or a review of your submission.
No. Certification makes you eligible to be considered for matching projects, but placement depends on project requirements and availability, and the terms rule out any guarantee of an offer.
It may use de-identified interview data to train its own models, says it does not share candidate data with AI model companies for their training, and offers an opt-out.
