Skip to content

Research · Compiled from public sources

Robot Training Data Vendors: Who Sells It, to Whom, and What They Pay

24 companies sell robot, embodied-AI, AV or drone training data, or the human work behind it; 57 facts about them are on record, each linked to its source.

Covers
28 Aug 2024 to 18 Oct 2026
Dated by when each event occurred
Calculated
18 Oct 2026, 16:34 UTC
Sources last checked 18 Oct 2026
Where it comes from
The vendors' own pages and announcements, filings and reporting by major outlets, each linked beside its fact with a verbatim quote kept in the dataset.
Method
Revision 1. Terms defined
CC BY 4.0, citing this report
Every figure as CSVJSON

What the data shows

  1. 1

    micro1 is the most documented vendor (5 sourced facts).

    Figure 1
  2. 2

    10 of 24 vendors name at least one robotics customer in public: micro1, Scale AI, DoorDash, Encord and Moove and others.

    Figure 1
  3. 3

    5 published pay figures from 4 vendors, from per-minute video capture to hourly annotation.

    Figure 2
  4. 4

    The most recent launch on record: DoorDash (19 Mar 2026): DoorDash launched Tasks, a stand-alone app (plus tasks inside the Dasher app) that pays its delivery couriers to film everyday tasks and record speech to improve AI and robotic systems, in select US places.

    Figure 3

Figure 1

The vendors and what each sells for robotics

Which companies sell robot training data or the human work behind it, and what does each offer?

VendorWhat it sellsNamed customersSourced facts
micro1TechCrunchBuilding a robotics pre-training dataset by having hundreds of generalists record everyday object interactions in their own homes (as CEO Ali Ansari told TechCrunch).1
5
ClaruClaru (company site)Claru, the training-data platform of Reka AI, runs capture infrastructure, enrichment pipelines and an annotation workforce to deliver datasets for robotics and physical AI (egocentric video, robotics trajectories, world-model footage).0
4
Scale AIScale AIScale AI's Physical AI Data Engine page says more than 1,000 hours of demonstration data are collected and uploaded to its platform every day, from data factories, at-home collectors and in-field commercial deployments, including teleoperated bimanual robots and a robotless egocentric rig (Scale Harness).1
4
Build AIBuild AI (company site)Build AI calls itself the data hyperscaler for physical AI: it makes its own capture hardware and runs collection of egocentric video of physical labor, and offers labs an off-the-shelf dataset with quality guarantees.0
3
DoorDashTechCrunchDoorDash launched Tasks, a stand-alone app (plus tasks inside the Dasher app) that pays its delivery couriers to film everyday tasks and record speech to improve AI and robotic systems, in select US places.1
3
EncordEncord via PR Newswire (corrected release, Bolsamania reprint)Encord said it works with more than 300 AI teams, naming Woven by Toyota, Skydio and AXA Financial among them (the corrected version of its Series C release).1
3
MercorMercor (job listing)Posted a US-based contributor listing to collect first-person video of everyday household activities for a leading physical AI company (listing created June 24 2026, now closed).0
3
MooveMoove (company newsroom)Moove raised a 250 million US dollar Series C at a 2.1 billion US dollar valuation, led by Mubadala and co-led by Woven Capital (Toyota) and Ion Pacific, to expand its autonomous vehicle fleet operations business.1
3
TolokaTolokaToloka released HomER (Home Egocentric Robotics), an open-source dataset of head-mounted egocentric videos of household tasks for robotics training, built on its self-service platform, covering 17 task categories and published on Hugging Face.0
3
VayVay (careers page)Vay pays Remote Drivers in Las Vegas 18 US dollars an hour during training and 25 US dollars an hour after training, plus a monthly safety-related bonus.1
3
AppenAppenAppen sells egocentric data collection for robots and embodied systems: stereo first-person capture with synchronized wrist pose and gripper state, alongside robotics demonstration, sensor fusion, LiDAR and autonomous vehicle data, captured in its own controlled kitchen, workshop, assembly and logistics facilities.0
2
iMeritThe Robot ReportCarbon Robotics' CTO said iMerit has labeled about a million plant images for the LaserWeeder's AI since the partnership began in 2020, data behind Carbon's large plant model.1
2
KognicKognic (company site)Kodiak AI, the autonomous truck developer, chose Kognic's sensor-fusion data labeling platform to scale its annotation pipelines.2
2
ObjectwaysObjectwaysObjectways sells egocentric and teleoperation data collection services and embodied AI datasets for training robots.0
2
PerlePerle (job posting on Rippling)Recruits egocentric video contributors (phone on a headstrap, at home and at work) to build data that teaches AI and robotics how people move; postings cover the US, Canada and Latin America, Egypt, and Indonesia and Malaysia; pay is per hour of accepted footage, varying by region.0
2
ProlificProlific (company site)Offers commissioned embodied-AI training data from its participant pool: egocentric video, manipulation task recordings and and physical scene annotation.1
2
Sensei RoboticsY Combinator (Launch YC post by Sensei)Sensei builds and runs a network of trained contractors (Senseis) who get task descriptions from robotics researchers and engineers and are paid to record demonstrations that fulfil them.0
2
TSMGTSMG (Lever job board)TSMG staffs on-site contract Robot Teleoperators (for example in Milpitas, California) who operate robots through teleoperation consoles and capture operational data for client robotics and AI programs.0
2
UberPYMNTSUber AI Solutions, Uber's AI data services business, acquired Segments.ai, a multi-sensor (lidar and camera) labeling platform for robotics and autonomous driving, to build out its lidar and multi-sensor annotation services.0
2
Deepen AIEIN Presswire (Deepen AI press release)TIER IV, the Autoware founder, expanded its use of Deepen AI's sensor-fusion annotation tools, including for robotaxi perception, after using Deepen's annotation suite for more than four years.1
1
SamaSamaSama sells training data annotation for robotics and manufacturing models, from assembly lines to autonomous mobile robots, with workflows that load up to 14 sensors (cameras, LiDAR, radar, ultrasonic) and multiple point clouds.0
1
TransdevTransdev North America (careers site)Transdev's Alternative Services division sells AV operations work to autonomous vehicle companies: L2 Operators work directly with self-driving vehicles on the road, troubleshoot software and hardware issues and feed back to engineering teams, and L3 Specialists test unreleased software and do mapping.0
1
TransPerfect DataForceTransPerfect DataForce (careers site)DataForce staffs on-site Tech Annotators in Paris who teleoperate robotic arms and robotic systems in real time and record and annotate the sessions, alongside AI prompt and evaluation work.0
1
Welo DataWelo DataWelo Data sells fully managed robotics and physical AI data collection: it rents the space, recruits participants, secures the site, runs collection, checks quality and delivers model-ready data.0
1

Base: 24 vendors, 57 facts, each linked to the page that states it (reviewed 18 Oct 2026).

A vendor appears once it has at least one sourced fact about its robotics work; a short row means little has been published, not that the company is small.

Credit: compiled by RobotTraining.jobs Research, Robot Training Data Vendors: Who Sells It, to Whom, and What They Pay, checked 18 Oct 2026.

CSV

Figure 2

What vendors say they pay the people who produce the data

What do robot data vendors and paid programs pay their collectors, operators and annotators?

VendorPay as stated
ClaruClaru (job posting)Egocentric household tasks capture in Latin America pays a flat 6 US dollars per hour of accepted footage.
ClaruClaru (job posting)Household chores first-person video capture in the United States pays a flat 15 US dollars per hour, paid by ACH every two weeks.
micro1Forbes Australia (Forbes, by Anna Tong)CEO Ali Ansari posted on X that micro1 pays 50 to 90 US dollars per hour for annotating videos of robots performing tasks.
ObjectwaysKhaleej Times (AFP)A Chennai homemaker who sends chore recordings to Objectways through its app told AFP the work pays 250 rupees an hour.
VayVay (careers page)Vay pays Remote Drivers in Las Vegas 18 US dollars an hour during training and 25 US dollars an hour after training, plus a monthly safety-related bonus.

Base: 5 pay figures, each as its source words it.

A vendor's stated rate is an offer, not measured earnings; the job boards this site reads show what open roles advertise.

Credit: compiled by RobotTraining.jobs Research, Robot Training Data Vendors: Who Sells It, to Whom, and What They Pay, checked 18 Oct 2026.

CSV

Figure 3

Every sourced fact

What exactly has each vendor or its press coverage said about its robotics data business?

VendorKindFact
AppenAppenWhat it sellsAppen sells egocentric data collection for robots and embodied systems: stereo first-person capture with synchronized wrist pose and gripper state, alongside robotics demonstration, sensor fusion, LiDAR and autonomous vehicle data, captured in its own controlled kitchen, workshop, assembly and logistics facilities.
AppenAppenWorkforceAppen says task demonstrations for its physical AI datasets are sourced from its global contributor network of more than 1 million people in more than 170 countries.
Build AIBuild AI (company site)What it sellsBuild AI calls itself the data hyperscaler for physical AI: it makes its own capture hardware and runs collection of egocentric video of physical labor, and offers labs an off-the-shelf dataset with quality guarantees.
Build AIBuild AI (company site)FundingBuild AI says it has raised more than 30 million US dollars, naming HF0, Abstract, Pear, Renegade and Costanoa among its backers.
Build AIHumanoids DailyWorkforceBuild AI's Egocentric-100K dataset was captured by 14,228 factory workers wearing its head-mounted cameras.
ClaruClaru (company site)What it sellsClaru, the training-data platform of Reka AI, runs capture infrastructure, enrichment pipelines and an annotation workforce to deliver datasets for robotics and physical AI (egocentric video, robotics trajectories, world-model footage).
ClaruClaru (company site)WorkforceOperates a data collection network of more than 10,000 trained collectors with wearable cameras across more than 100 cities on 6 continents.
ClaruClaru (job posting)Pay to workersEgocentric household tasks capture in Latin America pays a flat 6 US dollars per hour of accepted footage.
ClaruClaru (job posting)Pay to workersHousehold chores first-person video capture in the United States pays a flat 15 US dollars per hour, paid by ACH every two weeks.
Deepen AIEIN Presswire (Deepen AI press release)Named customerTIER IV, the Autoware founder, expanded its use of Deepen AI's sensor-fusion annotation tools, including for robotaxi perception, after using Deepen's annotation suite for more than four years.
DoorDashTechCrunchLaunchDoorDash launched Tasks, a stand-alone app (plus tasks inside the Dasher app) that pays its delivery couriers to film everyday tasks and record speech to improve AI and robotic systems, in select US places.
DoorDashTechCrunchWorkforceDoorDash Tasks can draw on more than 8 million Dashers across the US, according to its general manager Ethan Beatty.
DoorDashThe Next WebNamed customerWaymo is a Tasks customer: DoorDash couriers are paid to close open doors on Waymo self-driving cars, a program both companies confirmed and that sits inside the Tasks platform.
EncordEncord via PR Newswire (corrected release, Bolsamania reprint)Named customerEncord said it works with more than 300 AI teams, naming Woven by Toyota, Skydio and AXA Financial among them (the corrected version of its Series C release).
EncordSiliconANGLEFundingEncord raised a 60 million US dollar Series C led by Wellington Management to build data infrastructure for physical AI (robots, drones, autonomous vehicles), bringing its total funding to 110 million US dollars.
EncordEncord via PR Newswire (corrected release, Bolsamania reprint)RevenueEncord said its revenue from physical AI customers grew tenfold in the twelve months before its February 2026 Series C, while data on its platform grew from 1 petabyte to more than 5 petabytes.
iMeritThe Robot ReportNamed customerCarbon Robotics' CTO said iMerit has labeled about a million plant images for the LaserWeeder's AI since the partnership began in 2020, data behind Carbon's large plant model.
iMeritExlService Holdings (SEC Form 8-K, Exhibit 99.1)FundingExlService Holdings (EXL) agreed to acquire iMerit, a data annotation and model evaluation company serving robotics and autonomous mobility, for up to 310 million US dollars (170 million upfront plus up to 140 million in incentives and earnouts); EXL completed the acquisition on August 3, 2026.
KognicKognic (company site)Named customerKodiak AI, the autonomous truck developer, chose Kognic's sensor-fusion data labeling platform to scale its annotation pipelines.
KognicKognic (company site)Named customerKognic says its sensor-fusion annotation platform is used by Qualcomm, Bosch, Continental and Zenseact.
MercorTechCrunchRevenueFounder and CEO Brendan Foody said on X that the annualized revenue run rate crossed 2 billion US dollars.
MercorMercor (job listing)What it sellsPosted a US-based contributor listing to collect first-person video of everyday household activities for a leading physical AI company (listing created June 24 2026, now closed).
MercorTechCrunchFundingRaised 350 million US dollars at a 10 billion dollar valuation in a Series C led by Felicis Ventures (confirmed by the company).
micro1Forbes Australia (Forbes, by Anna Tong)FundingRaised a round of more than 100 million US dollars at a 4 billion dollar valuation (per two people familiar with the deal; micro1 declined to comment).
micro1Forbes Australia (Forbes, by Anna Tong)Named customerCounts robotics company 1X among its customers, alongside frontier AI labs, Microsoft and Amazon.
micro1Forbes Australia (Forbes, by Anna Tong)Pay to workersCEO Ali Ansari posted on X that micro1 pays 50 to 90 US dollars per hour for annotating videos of robots performing tasks.
micro1TechCrunchRevenueGross annual run rate grew from 100 million to 500 million US dollars over eight months (per a person familiar with the company).
micro1TechCrunchWhat it sellsBuilding a robotics pre-training dataset by having hundreds of generalists record everyday object interactions in their own homes (as CEO Ali Ansari told TechCrunch).
MooveMoove (company newsroom)FundingMoove raised a 250 million US dollar Series C at a 2.1 billion US dollar valuation, led by Mubadala and co-led by Woven Capital (Toyota) and Ion Pacific, to expand its autonomous vehicle fleet operations business.
MooveMoove (company newsroom)WorkforceMoove said its autonomous vehicle workforce was about 150 employees in August 2026 and that it expects to grow it to about 500 by the end of 2026.
MooveMoove (company newsroom)Named customerWaymo signed Moove as its fleet operations partner: Moove manages and dispatches Waymo's fully autonomous fleet and runs fleet operations, facilities and charging, starting in Phoenix in 2025 and expanding to Miami in 2026.
ObjectwaysKhaleej Times (AFP)Pay to workersA Chennai homemaker who sends chore recordings to Objectways through its app told AFP the work pays 250 rupees an hour.
ObjectwaysObjectwaysWhat it sellsObjectways sells egocentric and teleoperation data collection services and embodied AI datasets for training robots.
PerlePerle (job posting on Rippling)What it sellsRecruits egocentric video contributors (phone on a headstrap, at home and at work) to build data that teaches AI and robotics how people move; postings cover the US, Canada and Latin America, Egypt, and Indonesia and Malaysia; pay is per hour of accepted footage, varying by region.
PerleVenture Capital Access Online (Perle press release)FundingClosed a 9 million US dollar seed round led by Framework Ventures, bringing total funding to 17.5 million dollars.
ProlificProlific (company site)What it sellsOffers commissioned embodied-AI training data from its participant pool: egocentric video, manipulation task recordings and and physical scene annotation.
ProlificProlific (company site)Named customerSays it is trusted by more than 100 robotics and physical AI research teams.
SamaSamaWhat it sellsSama sells training data annotation for robotics and manufacturing models, from assembly lines to autonomous mobile robots, with workflows that load up to 14 sensors (cameras, LiDAR, radar, ultrasonic) and multiple point clouds.
Scale AIScale AIWhat it sellsScale AI's Physical AI Data Engine page says more than 1,000 hours of demonstration data are collected and uploaded to its platform every day, from data factories, at-home collectors and in-field commercial deployments, including teleoperated bimanual robots and a robotless egocentric rig (Scale Harness).
Scale AIUniversal RobotsLaunchUniversal Robots unveiled the UR AI Trainer, developed with Scale AI, which pairs UR robots with Scale AI software to capture imitation learning data on robots in production; the two companies said they would release a large industrial dataset collected on UR robots.
Scale AIScale AILaunchScale AI announced the expansion of its Data Engine to physical AI: a combined data collection and annotation service that supplies robotics companies with the datasets they need to train foundation models, extending its earlier autonomous vehicle data work to robotics.
Scale AIScale AINamed customerScale AI said it provides and enriches robotics data for Physical Intelligence, Generalist AI and Cobot, after more than 100,000 production hours completed at its San Francisco prototyping lab with help from contributors worldwide.
Sensei RoboticsY Combinator (Launch YC post by Sensei)LaunchSensei launched on Launch YC as a robotics training data service built on human demonstrations, saying it collects that data at a tenth of the cost and twice the speed of current approaches.
Sensei RoboticsY Combinator (Launch YC post by Sensei)What it sellsSensei builds and runs a network of trained contractors (Senseis) who get task descriptions from robotics researchers and engineers and are paid to record demonstrations that fulfil them.
TolokaTolokaWhat it sellsToloka released HomER (Home Egocentric Robotics), an open-source dataset of head-mounted egocentric videos of household tasks for robotics training, built on its self-service platform, covering 17 task categories and published on Hugging Face.
TolokaTolokaWorkforceToloka said its platform matched the HomER egocentric video collection task to more than 1,800 active contributors, drawn from a verified crowd spanning over 100 countries and 40 languages.
TolokaSiliconANGLEFundingToloka closed a 72 million US dollar investment led by Bezos Expeditions, with Shopify CTO Mikhail Parakhin participating; parent Nebius gave up majority voting control.
TransdevTransdev North America (careers site)What it sellsTransdev's Alternative Services division sells AV operations work to autonomous vehicle companies: L2 Operators work directly with self-driving vehicles on the road, troubleshoot software and hardware issues and feed back to engineering teams, and L3 Specialists test unreleased software and do mapping.
TransPerfect DataForceTransPerfect DataForce (careers site)What it sellsDataForce staffs on-site Tech Annotators in Paris who teleoperate robotic arms and robotic systems in real time and record and annotate the sessions, alongside AI prompt and evaluation work.
TSMGTSMG (Lever job board)What it sellsTSMG staffs on-site contract Robot Teleoperators (for example in Milpitas, California) who operate robots through teleoperation consoles and capture operational data for client robotics and AI programs.
TSMGTSMG (Lever job board)What it sellsTSMG recruits participants (for example in Pattaya, Thailand) to wear a head-mounted camera while doing everyday or work tasks for a large-scale robotics AI video data collection project, recording at least 4 hours of video per project day.
UberTechCrunchLaunchUber launched an AV Labs division that sends its own sensor-equipped cars into cities to collect driving data for its autonomous vehicle partners (more than 20 at launch, including Waymo, Waabi and Lucid Motors).
UberPYMNTSWhat it sellsUber AI Solutions, Uber's AI data services business, acquired Segments.ai, a multi-sensor (lidar and camera) labeling platform for robotics and autonomous driving, to build out its lidar and multi-sensor annotation services.
VayVay (careers page)Pay to workersVay pays Remote Drivers in Las Vegas 18 US dollars an hour during training and 25 US dollars an hour after training, plus a monthly safety-related bonus.
VayVay (company newsroom)FundingGrab agreed to invest 60 million US dollars in remote-driving company Vay, with a potential further 350 million US dollars if financial and operational milestones are met within the first year after closing.
VayTechCrunchNamed customerKodiak Robotics uses Vay's remote-driving system for its self-driving trucks: CDL-licensed operators remotely drive the trucks in low-speed, defined scenarios while Kodiak's system keeps running its checks.
Welo DataWelo DataWhat it sellsWelo Data sells fully managed robotics and physical AI data collection: it rents the space, recruits participants, secures the site, runs collection, checks quality and delivers model-ready data.

Base: All 57 facts. "Reported" means an outlet reports it from sources the company has not confirmed.

Amounts stay in the currency and form their source gives; nothing is converted or totalled.

Credit: compiled by RobotTraining.jobs Research, Robot Training Data Vendors: Who Sells It, to Whom, and What They Pay, checked 18 Oct 2026.

CSV

Robotics labs · Fleet operators · Data vendors · Research teams

Buy the data behind these reports

employer job boards read every day
103employer job boards read every day
every role's history kept since
12 Sep 2026every role's history kept since
apply clicks in the last 30 days
392apply clicks in the last 30 days
  • The raw records

    Every robot training role we track: title, employer, place, kind of work, pay as stated, and the days it was first and last open.

  • A regular feed

    New and closed roles every week or month, as file drops or through the data endpoints.

  • A custom cut

    Any figure here by country, city, kind of work, vehicle, employer or date, or your own hiring benchmarked against the market.

Each report's own figures stay free to download and reuse under CC BY 4.0.

How it was measured

The vendors' own pages and announcements, filings and reporting by major outlets, each linked beside its fact with a verbatim quote kept in the dataset. Each row carries a link to its publication and gives the number exactly as published, with no estimates or conversions. A note under every figure says what it covers.

Vendor
A company that sells robot training data, or the human work that produces it, to companies building robots, embodied AI, autonomous vehicles or drones.
Stated or reported
Stated: the company, a filing or an official page says it. Reported: an outlet reports it from sources the company has not confirmed.

Limits of this data

  • Built only from public sources. Anything never announced or reported is absent, and private-company figures are whatever the company or the press said, unaudited.
  • Vendors publish unevenly, so the number of facts reflects what is public, not a vendor's size.
  • Pay figures are the vendors' own stated offers, not measured earnings.

Collection and classification of listings are explained on the methodology page. Freshness, completeness and consistency are tested daily in the data quality report.

Monthly editions

The live page shows the dataset as of its latest review, and new events join it once they are made public. Every finished month also gets its own edition, locked two days after month end, so a number quoted from it stays fixed.

Quoting this report

You may quote, chart or republish these figures under CC BY 4.0, provided you link back to this page. The live numbers shift every day. For a figure that stays put, quote a monthly edition, or note the date you accessed this page.

Citation

RobotTraining.jobs (2026). Robot Training Data Vendors: Who Sells It, to Whom, and What They Pay (live data, 28 Aug 2024 to 18 Oct 2026). https://robottraining.jobs/research/robot-training-data-vendors (accessed 18 Oct 2026)

BibTeX

@misc{robottrainingjobs_robot_training_data_vendors_live,
  author = {{RobotTraining.jobs}},
  title = {Robot Training Data Vendors: Who Sells It, to Whom, and What They Pay (live data, 28 Aug 2024 to 18 Oct 2026)},
  year = {2026},
  howpublished = {\url{https://robottraining.jobs/research/robot-training-data-vendors}},
  note = {Accessed 18 Oct 2026},
}

Files: every figure in one CSV (a row per value) or as JSON. Each figure can also be downloaded separately. Other open data: datasets and API.

Keep reading

Browse all 68 reports