What if the most valuable thing you do all day isn’t your job, but the way you make your morning coffee? That’s the bet behind Mecka AI, a two-year-old startup that just raised a $60 million Series B led by Sequoia, with Nvidia and Microsoft’s venture fund M12 joining in. Mecka pays people to record themselves doing ordinary tasks, then turns that footage into training data for humanoid robots and other machines.
TechCrunch reported the round on October 7, 2026, and described the company’s goal as doing for robotics what data-labeling firms like Scale AI did for chatbots.
So why would Sequoia and Nvidia back a company that films people cooking dinner? Let’s break it down.
| Key fact | Details |
|---|---|
| Company | Mecka AI (also styled Mecka) |
| What it does | Collects and processes real-world human activity data to train robots |
| Latest round | $60 million Series B, announced October 7, 2026 |
| Lead investor | Sequoia Capital |
| Other backers | Nvidia, Microsoft’s M12, Qualcomm’s venture arm, Samsung, Kindred Ventures, Framework Ventures, Neo |
| Earlier funding | $60 million Series A in two parts ($25M + $35M) |
| Reported valuation | About $500 million, per earlier TechCrunch reporting (not confirmed) |
| CEO | Josh Gao, co-founder |
| Offices | Toronto and New York City |
| Staff | About 45, most of them in Toronto (as of June) |
What is Mecka AI, in simple terms?
Mecka calls itself “the data and deployment layer for physical AI.” Mecka doesn’t build robots. It captures how real people move through real tasks, cleans and labels that footage, and sells the result to robotics labs and big tech companies that are training robot brains.
Contributors wear body sensors and use iPhones to record first-person video while they work. According to BetaKit, that recording happens in homes, kitchens, chemistry labs, metal fabrication shops and even leather workshops. Mecka’s own hardware and computer vision models then convert the raw clips into data a robot model can learn from.
On its website, Mecka is blunt about its role: “We don’t build robots — we’re the integrator.” It also offers a mobile app for collection, data tools for browsing and querying datasets, and an enterprise service for evaluating real-world robot deployments.
The key takeaway here is simple: Mecka sells the lessons, not the robot.
Why robots need recordings of you
Here’s the thing. Chatbots like ChatGPT learned from billions of pages of text that already existed online. Robots don’t have that luxury. There’s no internet archive showing exactly how a hand grips a mug or tightens a bolt.
BetaKit put it neatly when it noted that this kind of material “cannot be scraped from the internet.” The Logic made the same point, saying robot makers “can’t just scrape the web for examples of how their machines should function.”
That’s where first-person video comes in. Film a task from your own point of view, add movement sensor data, and you get something close to a step-by-step lesson in physical work.
Does it work? Early research suggests it helps. A RuntimeWire analysis of the EgoVerse research paper found that co-training robots with human data generally improved their performance, especially when the demonstrators and scenes were varied.
“Chatbots learned from what people wrote. Robots will learn from what people do.”
If you’ve been following the rise of physical AI, this is the missing piece. The bigger picture is laid out in this look at why 2026 is the ChatGPT moment for robotics, and data was always the bottleneck in that story.
How Mecka AI collects its data
Let’s back up a little and look at the actual process. Based on what Mecka and the outlets covering it have published, the pipeline looks like this:
- Capture: Contributors record everyday tasks, such as making coffee, moving objects or repairing cars, using smartphones and wearable body sensors.
- Process: Mecka’s internal video understanding lab applies motion tracking and 3D reconstruction to the footage.
- Annotate and check: The company says teams need “annotation, context, QA, motion understanding, and evaluation workflows” before raw video becomes useful.
- Deliver: Clean datasets go to robotics labs and big tech customers, who can also commission task-specific collection.
Mecka is also an industry partner in EgoVerse, a shared human-data project for robot learning. The EgoVerse site lists 4,003 hours of human demonstrations, about 80,000 episodes, 1,965 tasks and 2,087 unique demonstrators, built by a consortium that includes Georgia Tech, Stanford, Meta and Scale AI alongside Mecka.
The key point: this isn’t one company filming in a lab. It’s a network of real people in real places.
Can you get paid to record yourself for Mecka?
This is the question most people searching for Mecka AI want answered. TechCrunch reports that Mecka “pays people to record themselves doing everyday tasks, like making coffee or fixing cars.” So yes, contributors are paid.
Here’s the catch. None of the published reports say how much contributors earn, how you sign up, or which countries are eligible. If you’re interested, check Mecka’s official site and read any contributor terms first.
The takeaway: paid work exists, but the details aren’t public yet.
The Mecka AI funding story so far
Mecka’s money has arrived fast. Here’s how the rounds stack up, based on BetaKit, AI Business, RuntimeWire and TechCrunch reporting.
| Round | Amount | Timing | Led by | Notable participants |
|---|---|---|---|---|
| Seed | $8 million (reported) | August 2025 | Not disclosed | Not disclosed |
| Series A | $25 million | November 2025 | Framework Ventures | SV Angel, Ted Xiao |
| Series A extension | $35 million | Closed by June 2026 | Framework Ventures | Menlo Ventures, Kindred Ventures |
| Series B | $60 million | Announced October 7, 2026 | Sequoia Capital | Nvidia, Microsoft’s M12, Qualcomm, Samsung, Neo |
BetaKit also reports that individual investors in the Series B include DoorDash CEO Tony Xu, former Snowflake and ServiceNow CEO Frank Slootman, and Milan Kovac, the former head of Tesla’s Optimus program. When a former humanoid program chief writes a check, it says a lot about where the hard part is.
What about revenue? BetaKit and The Logic report that Mecka says it passed a $100 million annual run rate in June and expects to hit $300 million by the end of 2026. Treat those as company claims, since Mecka hasn’t named a single customer.
The new money goes toward data infrastructure, Mecka’s research lab and a robot deployment layer.
The key takeaway: investors are betting on the picks-and-shovels side of the robot boom.
Who founded Mecka AI and where is it based?
Mecka has four co-founders, according to RuntimeWire: Josh Gao, Mogen Cheng, Jason Chong and Duy Nguyen. Gao, the CEO, and Cheng previously built restaurant payments technology and sold that business in 2023. Chong was a Microsoft software engineer who co-founded Utopia Labs, which Coinbase acquired. Nguyen came from the sneaker resale world and later ran Diamond Agency.
None of them came from classic robotics, but they know how to build marketplaces and operations at scale.
The company works out of Toronto and New York City, though it’s legally domiciled in the US. BetaKit reported in June that 40 of its 45 staff were in Toronto, and that three of the four co-founders are Canadian. Mecka also bought Docula, a small Vancouver startup, a few months before the Series B.
Gao summed up the company’s realism in a June news release: “Robotic deployment use cases are messy in the real world and not all tasks make sense today.” — Josh Gao, Mecka co-founder and CEO, via BetaKit
Where Mecka fits in the humanoid robot race
Tesla’s Optimus, Figure’s robots and Boston Dynamics’ Atlas grab the headlines. But every one of those machines needs huge amounts of training data to cope with the world outside a lab.
That’s the gap Mecka is chasing. It’s not competing with robot makers. It wants to sell to them.
This breakdown of the Boston Dynamics Atlas hand shows how much dexterity matters, and dexterity is exactly what human hand-motion data teaches. The piece on whether Tesla Optimus Gen 3 can fold your laundry gets at the same question from the household side.
Mecka isn’t alone in the data race, either. Here’s how it compares with others mentioned in recent coverage.
| Company | Approach | Latest reported funding news |
|---|---|---|
| Mecka AI | Human first-person video and body sensor data | $60M Series B led by Sequoia; about $500M valuation reported |
| XDOF | Robot data tied to teleoperation roots | Reportedly in talks at about a $1.2B valuation, per TechCrunch |
| Scale AI | Human data provider expanding from LLMs into robotics | Moving into physical AI, per TechCrunch |
| Micro1 | Human data provider expanding into robotics | Moving into physical AI, per TechCrunch |
Fast forward a few years and the winners in robotics might not be the companies with the shiniest robot. They might be the ones with the deepest, most varied library of human movement.
“In the robot race, the real currency may be footage of ordinary people doing ordinary things.”
The privacy questions you should ask
Recording first-person video inside homes and workplaces raises real privacy questions. A family member walking past or the paperwork on your desk can easily end up in frame.
What does Mecka say? Its privacy policy, effective March 20, 2026, covers “dataset contributions” and says, “We do not sell your personal information.” It lists rights such as access, correction, deletion and withdrawing consent, depending on where you live.
That said, the policy doesn’t specifically mention video, location or biometric data, and it doesn’t explain how contributor footage is shared with the companies that buy datasets. None of the news coverage addressed consent or data handling either.
If you ever sign up to record for Mecka or a similar company, a few steps help protect you:
- Check what’s in frame: Clear away mail, screens and photos before you start recording.
- Ask about blurring: Find out whether faces and bystanders are anonymized before data reaches customers.
- Read the deletion terms: Know whether you can pull your footage after it’s been sold or used in training.
- Keep others out: Don’t film family members or coworkers who haven’t agreed to it.
The bottom line on privacy: the rules for this new kind of gig work are still being written, so ask questions first.
What comes next for Mecka and robot training data
For years, robots were stuck in tightly controlled factory cells because they couldn’t cope with surprises. Companies like Mecka want to change that by feeding them the messy variety people handle without thinking. You can already see early steps, like the humanoid pilots inside BMW and Xiaomi factories.
With Sequoia, Nvidia and Microsoft’s M12 now on the cap table, Mecka has the money and the connections to scale fast. Whether its revenue claims hold up, and whether it handles contributor privacy well, will decide if it becomes the Scale AI of robotics or just one supplier among many.
Either way, the next time you flip an egg or tighten a screw, remember: that simple motion might be worth more than you think.
Source: Mecka AI’s official site.
Frequently asked questions
Mecka AI collects real-world human activity data, mostly first-person video and body sensor readings, and turns it into training data for robots. It describes itself as the data and deployment layer for physical AI.
Mecka was co-founded by Josh Gao, Mogen Cheng, Jason Chong and Duy Nguyen, according to RuntimeWire. Gao serves as CEO. The founders come from payments, software engineering and e-commerce backgrounds rather than traditional robotics.
Nvidia joined Mecka AI’s $60 million Series B in October 2026, alongside lead investor Sequoia and Microsoft’s M12.
Mecka operates out of Toronto and New York City, and it’s legally domiciled in the United States. BetaKit reported in June 2026 that 40 of its 45 employees were based in Toronto.
Yes. TechCrunch reports that Mecka pays people to record everyday tasks like making coffee or fixing cars, using smartphones and body sensors. Pay rates and sign-up details haven’t been published, so check Mecka’s official site for current information.
TechCrunch reported in September 2026 that Mecka was raising at about a $500 million valuation, but neither Mecka nor the October coverage confirmed the final figure. Treat that number as reported, not official.
It depends on the company’s terms. Mecka’s privacy policy says it doesn’t sell personal information and lists deletion and consent rights, but it doesn’t specifically address video or biometric data. Clear private items from view and read the contributor terms before you record.
