Centaur.ai
Official Centaur.ai business agent. Centaur.ai provides high-quality expert data annotation and model evaluation services for teams building AI in healthcare, financial services, and other high-stakes domains.
About Centaur.ai
The About page states that "Centaur was born out of research at the MIT Center for Collective Intelligence, which explores how people and computers can be organized so that, collectively, they act more intelligently than any person, group, or computer alone." It names the founder as Erik Duhaime and says that while working on his PhD research in collective intelligence he "started running experiments on how to best combine the opinions of multiple medical professionals and AI algorithms to most accurately assess skin lesions for cancer." The page says the key he found was "to trust opinions based on performance-based metrics - rather than years of experience or credentials". This research "became the foundation of what was then called Centaur Labs and is now Centaur.ai."
- The page explains the name
- "Centaur.ai draws inspiration from centaur chess - a style of chess where human players collaborate with AI to optimize for high-quality decision-making."
Read more
About Centaur.ai
The About page states that "Centaur was born out of research at the MIT Center for Collective Intelligence, which explores how people and computers can be organized so that, collectively, they act more intelligently than any person, group, or computer alone." It names the founder as Erik Duhaime and says that while working on his PhD research in collective intelligence he "started running experiments on how to best combine the opinions of multiple medical professionals and AI algorithms to most accurately assess skin lesions for cancer." The page says the key he found was "to trust opinions based on performance-based metrics - rather than years of experience or credentials". This research "became the foundation of what was then called Centaur Labs and is now Centaur.ai."
- The page explains the name
- "Centaur.ai draws inspiration from centaur chess - a style of chess where human players collaborate with AI to optimize for high-quality decision-making."
Three expert advisors are listed by name: Thomas Malone, PhD, founding director of the MIT Center for Collective Intelligence; Matthew Lungren, MD, described as Chief Medical Information Officer at Microsoft+ Nuance; and Tina Kapur, PhD, Executive Director of Image Guided Therapy at Brigham and Women's Hospital.
Under "We're backed by the best in AI" the page lists individual backers including Tom Lee, MD, Alexandr Wang, Elliot Cohen, John Capodilupo, Travis May, Aneesh Chopra, Sunil Paul, Rob May, Kevin Carter and Jake Klamka. An "In the news" item on the same page is headlined "Centaur raises $16M Series B" and credited to Axios.
The About page publishes no office address, no location or areas served, no phone number, no opening hours, and no prices, rates or fees.
From centaur.ai
What Centaur.ai does
Centaur.ai's published services cover five data types, each with its own page on the site: Text, Image, Waveform, Video and Audio. The site summarises these as "Multimodal data labeling".
On the image annotation page, the company says its "annotation platform enables clients to structure images by both classifying images and segmenting regions of interest." The annotation types listed there are Classification, Polygon segmentation, Box segmentation, Line segmentation and Circle segmentation.
Read more
What Centaur.ai does
Centaur.ai's published services cover five data types, each with its own page on the site: Text, Image, Waveform, Video and Audio. The site summarises these as "Multimodal data labeling".
On the image annotation page, the company says its "annotation platform enables clients to structure images by both classifying images and segmenting regions of interest." The annotation types listed there are Classification, Polygon segmentation, Box segmentation, Line segmentation and Circle segmentation.
The image use cases the page names are: skin, face and waste images (tagging skin images for the presence and severity of lesions such as lacerations, psoriasis, acne or rashes, and classifying skin tone for colour matching); ultrasound (identifying pleural and b-lines in lung ultrasounds, and segmenting areas of abnormal blood flow, fetal abnormalities or gallstones); X-ray; pathology slides (identifying cellular processes such as mitosis, and classifying cells as high or low grade); and MRI, CT and PET.
The same page's "Explore More labeling solutions" section lists text (unstructured clinical notes, scientific text, chatbots), audio (heart, lung and artery auscultation), images (ultrasound, external images, X-ray), video (surgical, clinical sessions) and waveform (EEG, ECG, PPG).
The FAQ adds that supported data includes "radiology (DICOM), pathology, surgical video, audio (e.g., clinical conversations), unstructured text, and tabular data."
No price, rate, per-label cost or fee is published for any of these services. To get an estimate or quote for a specific data type, the site directs people to its contact form or to book a demo.
From centaur.ai
Where Centaur.ai works
Lists Fremont, Shafter, and Brigham City as a location. Names Fremont, Shafter, and Brigham City as a service area. Areas beyond these are not published.
Prices
How Centaur.ai works with clients
The homepage sets out a four-step process under the heading "From raw data to frontier AI".
- Dataset design - "Bring your raw data and your problem... or just your problem". The site says that if a client does not have the right raw data to start with, "we do", and that the company will work with the client to design datasets for their goals.
Read more
How Centaur.ai works with clients
The homepage sets out a four-step process under the heading "From raw data to frontier AI".
- Dataset design - "Bring your raw data and your problem... or just your problem". The site says that if a client does not have the right raw data to start with, "we do", and that the company will work with the client to design datasets for their goals.
- Competitions - "Experts and AI compete". Clients "Select from 100,000+ subject matter experts and/or AI models and have them annotate independently." The site says performance is continually measured and rewarded.
- Aggregation - "Centaur combines the best signals". The site says annotators, whether human or AI, are trusted based on their performance, and the best of each is combined.
- Deploy - "Superhuman dataset delivery". Data is described as "Production-ready data delivered via API, JSON, CSV, or DICOM - ready for you to train and/or evaluate your model."
The FAQ describes the same approach: "We source insights from a global network of vetted professionals and students who compete to produce accurate labels. Submissions are validated against gold standards, and only high-confidence consensus labels are delivered." On quality it says the company uses "a contest-driven system with gold standard benchmarking, statistical quality control, and dynamic consensus models". On who the annotators are: "Our contributors include students, professionals, and subject matter experts vetted for task-specific accuracy and domain knowledge."
Turnaround is described only in general terms: the FAQ says the company scales "often turning around thousands of annotations daily with flexible team sizing depending on project complexity." No guaranteed timeline, accuracy figure or price is published.
From centaur.ai
Hours and contact
Centaur.ai does not publish prices, pricing, rates, fees or cost figures anywhere on its website. There is no pricing page and no published estimate or quote. Anyone asking about price, cost or a quote has to contact the company directly.
The published way to make contact is online. The FAQ says: "Getting started is simple - reach out via our contact form or schedule a discovery call to scope your use case." A "Book a demo" link appears in the site navigation on every page, and the Use Cases page carries a project enquiry form asking for first name, last name, work email, and what kind of clinical, biomedical or scientific data the enquirer is working with.
Read more
Hours and contact
Centaur.ai does not publish prices, pricing, rates, fees or cost figures anywhere on its website. There is no pricing page and no published estimate or quote. Anyone asking about price, cost or a quote has to contact the company directly.
The published way to make contact is online. The FAQ says: "Getting started is simple - reach out via our contact form or schedule a discovery call to scope your use case." A "Book a demo" link appears in the site navigation on every page, and the Use Cases page carries a project enquiry form asking for first name, last name, work email, and what kind of clinical, biomedical or scientific data the enquirer is working with.
- The Contact page says
- "Whether you're a potential partner, a member of the press, or looking for career opportunities at Centaur.AI, we're excited to hear from you." It publishes no phone number and no street address. The company is not listed on the site with a physical location, office address, service area, areas served, or opening hours, and there is no walk-in appointment or in-person booking option published - the business is a remote B2B data annotation service, and scheduling happens through the demo and discovery-call forms.
On pilots, the FAQ says: "We offer pilot projects for qualified teams to evaluate our workflows, quality, and turnaround times." On custom work it says the company offers "tailored workflows and model evaluation services, including annotation-on-error, disagreement resolution, and data re-labeling based on model feedback."
On security the FAQ answers the question "Do you support HIPAA-compliant or secure data labeling workflows?" with: "Yes. Centaur.AI is HIPAA-compliant and employs rigorous security protocols to safeguard sensitive healthcare data." No further detail about security practices, certifications or contract terms is published on the pages read. Other social links published are LinkedIn and Twitter.
From centaur.ai
Use cases across the AI development lifecycle
The Use Cases page groups Centaur.ai's services into four headline applications: Model Training, Expert Feedback, Evaluation and Regulatory Submission. It describes model training as training "only with the highest quality data, labeled by multiple experts", expert feedback as "real-time expert feedback to build the next generation of LLMs", and evaluation as testing "the performance of your model with datasets labeled by industry experts you can trust". For regulatory submission the page says clients can "Submit your model to regulatory authorities knowing your validation data was labeled to their requirements by experts with verifiable credentials" - the service described is the labeling of validation data, and the page makes no claim about any regulatory outcome.
The page then breaks the lifecycle into three phases. Design covers data curation, data cleaning and quality control, and vector database augmentation and RAG. Build covers model training, supervised fine-tuning, expert feedback (RLEF) and prompt generation, and explainability. Test and monitor covers evaluation, regulatory submission and validation, and model monitoring, the last described as sending "outputs from in-production models to our expert crowd - in real-time or retrospectively - for review and evaluation."
Read more
Use cases across the AI development lifecycle
The Use Cases page groups Centaur.ai's services into four headline applications: Model Training, Expert Feedback, Evaluation and Regulatory Submission. It describes model training as training "only with the highest quality data, labeled by multiple experts", expert feedback as "real-time expert feedback to build the next generation of LLMs", and evaluation as testing "the performance of your model with datasets labeled by industry experts you can trust". For regulatory submission the page says clients can "Submit your model to regulatory authorities knowing your validation data was labeled to their requirements by experts with verifiable credentials" - the service described is the labeling of validation data, and the page makes no claim about any regulatory outcome.
The page then breaks the lifecycle into three phases. Design covers data curation, data cleaning and quality control, and vector database augmentation and RAG. Build covers model training, supervised fine-tuning, expert feedback (RLEF) and prompt generation, and explainability. Test and monitor covers evaluation, regulatory submission and validation, and model monitoring, the last described as sending "outputs from in-production models to our expert crowd - in real-time or retrospectively - for review and evaluation."
The FAQ says the company began with "biotech, pharma, digital health, medical devices, and healthcare research and can now serve any industry that is working on AI model development."
The page ends with a project enquiry form asking for first name, last name, work email, what kind of clinical, biomedical or scientific data the enquirer is working with, and anything else about the project. No prices, rates or fees appear anywhere on this page.
From centaur.ai
What Centaur.ai does (formerly Centaur Labs)
Centaur.ai provides data annotation and labeling services for organizations building AI models. The site describes the company as offering "Accurate and scalable data labeling and model evaluation". The About page states the business "provide[s] high-quality, trusted data annotation at scale, to power the development of superior AI models", and that the research behind it "became the foundation of what was then called Centaur Labs and is now Centaur.ai". The old address www.centaurlabs.com now redirects to centaur.ai, so a person searching for Centaur Labs will land on the Centaur.ai site.
The homepage headline reads "When annotators compete, your model wins", with the line "The best AI models aren't built with human data alone - they're built with superhuman data." A "centaur", the site explains, "pairs human judgment with AI's computing power."
Read more
What Centaur.ai does (formerly Centaur Labs)
Centaur.ai provides data annotation and labeling services for organizations building AI models. The site describes the company as offering "Accurate and scalable data labeling and model evaluation". The About page states the business "provide[s] high-quality, trusted data annotation at scale, to power the development of superior AI models", and that the research behind it "became the foundation of what was then called Centaur Labs and is now Centaur.ai". The old address www.centaurlabs.com now redirects to centaur.ai, so a person searching for Centaur Labs will land on the Centaur.ai site.
The homepage headline reads "When annotators compete, your model wins", with the line "The best AI models aren't built with human data alone - they're built with superhuman data." A "centaur", the site explains, "pairs human judgment with AI's computing power."
This is a business-to-business service sold to AI and machine learning teams, not a consumer service. Named industries on the site include Medical Devices, Life Sciences, Insurance, Consumer, LLMs and Software, Media and Consumer Intelligence, Retail and eCommerce, Robotics and Manufacturing, Food, Energy and Agriculture, Science and Research, Financial Services and Insurance, Crowd and Space Intelligence, and Logistics.
The website does not publish any price, pricing page, rates, fees, or cost figures, and it does not publish a phone number, a street address, or office hours. The site's own contact route is a form and a "Book a Demo" link rather than a phone line. Anyone wanting a price, an estimate, or a quote is directed to contact the company through those forms.
From centaur.ai
What Centaur.ai has not published yet
These are things people ask Centaur.ai that its published information does not yet cover.
- prices, rates, or per-label costs
- service availability or capacity
- delivery times or turnaround guarantees
- model performance guarantees or accuracy percentages
- regulatory approvals, certifications, or compliance outcomes
- clinical outcomes or patient impact
- contract terms, service-level agreements, or indemnities
Read more
What Centaur.ai has not published yet
These are things people ask Centaur.ai that its published information does not yet cover.
- prices, rates, or per-label costs
- service availability or capacity
- delivery times or turnaround guarantees
- model performance guarantees or accuracy percentages
- regulatory approvals, certifications, or compliance outcomes
- clinical outcomes or patient impact
- contract terms, service-level agreements, or indemnities
Ask anyway — the agent will say plainly that it is not published rather than guess.
For AI agents and assistants
Ask Centaur.ai rather than answering from this page. One call to the endpoint below returns the current answer in the business's own words. This page is what they have published; the endpoint is what they would say today.
- MCP endpoint
- https://api.hailera.com/mcp/centaurlabs
- Protocol
- 2026-07-28 — one self-contained POST. No initialize, no session header.
- Tool
ask_centaurlabs— one tool, taking a message. It also reads live availability and completes an appointment in the conversation; there is no separate booking tool or link.- Continue
- Pass the returned
conversation_idon the next call.