Human Archive

Multimodal data provider for robotics and world modeling

Winter 2026ActiveIndustrialsManufacturing and RoboticsArtificial IntelligenceRoboticsData LabelingSan Francisco, CA, USA

Company

https://www.humanarchive.ai

We’re archiving the physical world for embodied intelligence by collecting and labeling aligned multimodal data. To build dexterous and perceptive robots that generalize robustly, we need massive amounts of real-world data across multiple modalities and environments. We have thought deeply about the fine line between biomimicry and its application to humanoid systems. Based on this research, we design and deploy custom hardware across residential and manufacturing settings. We then post-process the resulting data through internal QA, anonymization, and annotation pipelines to deliver diverse, high-fidelity datasets at scale to frontier labs developing robotics foundation models and general-purpose robotics companies. We believe we are at a historic inflection point, with a unique opportunity to leave a dent on humanity and reshape physical labor markets forever. That's why our team dropped out of Stanford and Berkeley and moved to Asia to collect the world’s largest annotated multimodal dataset.

Verdict

High Signal

Market Opportunity

Robotics training data is a massive and fast-growing market as humanoid robot companies (Figure, 1X, Physical Intelligence, etc.) race to build foundation models and desperately need diverse real-world multimodal datasets. The ICP is clear: frontier robotics labs and general-purpose robotics companies. Monetization path via data licensing/sales is well understood. TAM is easily $1B+ given the scale of capital flowing into embodied AI.

Medium Signal

Founder Signal

Four young founders from Berkeley/Stanford with limited real work experience. Rushil (Berkeley MET) had a PM internship at Coinbase and a prior acquired startup with $25k MRR — the most substantive signal on the team. Samay (Berkeley EECS, on leave) did SDE at Amazon for 4 months and ML work at Lightning AI. Raj is a Berkeley dropout whose primary listed experience is farming/mango-selling for 9 years. Shloke is a current Stanford ME/CS researcher. Team is young and light on direct robotics data industry experience, though technical backgrounds exist.

Medium Signal

Competition

Competitors include Scale AI (dominant data labeling player with robotics focus), Apptronik data initiatives, Physical Intelligence's internal data collection, and other robotics data startups. The differentiation claim — custom hardware deployment in residential and manufacturing settings in Asia for diversity — is plausible but unverified. No proprietary moat is demonstrated yet; Scale AI has massive advantages in infrastructure and enterprise relationships.

Low Signal

Product

Website is essentially a single tagline with a contact email — no demos, no product screenshots, no pricing, no customer logos, no API docs. Zero visible evidence of data collection scale, labeling pipelines, or any delivered datasets. Pure vaporware presentation at this stage.

OverallC Tier

The market thesis is legitimately strong — robotics training data is one of the hottest infrastructure gaps in AI right now — but this team has not demonstrated execution at any meaningful scale. The website is effectively blank, there are no disclosed customers, no data on how many hours or environments have been collected, and the founders are predominantly very recent students or dropouts with thin relevant experience. The claim of moving to Asia to collect the world's largest annotated multimodal dataset is a bold narrative but completely unverified. Rushil's prior acquisition and $25k MRR gives some signal that at least one founder can ship, but the rest of the team needs to prove they can actually deliver the hardware deployment and data pipeline they're describing at scale before this becomes compelling.