Humyn Labs

Research and data platform capturing human activity for Physical AI training datasets.

Website: https://humynlabs.ai/

Cover Block

Public sources

Field Detail
Name Humyn Labs
Tagline Research and data platform capturing human activity for Physical AI training datasets.
Headquarters Bengaluru, India [TechCrunch, November 2026]
Founded 2026 [TechCrunch, November 2026]
Stage Series A [TechCrunch, November 2026]
Business model B2B
Industry Deeptech
Technology AI / Machine Learning
Geography Global / Remote-First
Growth profile Venture Scale
Founding team Co-Founders (2), Manish Agarwal and Ishank Gupta [TechCrunch, November 2026]
Funding label $50M+
Total disclosed ~$47,000,000 [TechCrunch, November 2026]

Links

Public sources

Executive Summary

PUBLIC Humyn Labs is building a research and data platform for Physical AI, and it has drawn investor attention quickly because it sits in a narrow but increasingly important part of the robotics stack: collecting real-world human activity data that can be turned into training sets for robot and world-model systems [TechCrunch, November 2026] [Bloomberg, November 2026]. Founded in 2026 in Bengaluru by Manish Agarwal and Ishank Gupta, the company appears to have moved from formation to institutional financing at unusual speed, announcing a $47 million Series A package in November 2026 that combined $26 million of equity led by Accel and $21 million of venture debt led by Lighthouse Canton Growth Debt Fund II [TechCrunch, November 2026] [Business Standard, November 2026].

The product is best understood as an end-to-end data engine: Humyn records people performing tasks in homes, factories, warehouses, and farms, then validates, processes, and labels that material into training-ready datasets for robotics developers, frontier AI labs, and companies building world models [The Times of India, November 2026] [TechCrunch, November 2026]. Its differentiation, based on current public reporting, rests less on model IP and more on operating infrastructure, specifically a distributed contributor network, field collection capability in more than 20 countries, and company-developed two-camera hardware for capture in real environments [TechCrunch, November 2026] [Business Standard, November 2026].

The founding team brings operating experience rather than academic robotics credentials. Agarwal was previously CEO of Nazara Technologies, while Gupta is described by TechCrunch as a former executive at AB InBev, a mix that suggests execution depth and commercial orientation, though public evidence on domain-specific robotics research experience remains limited [TechCrunch, November 2026] [Forbes, March 2018].

The business model is B2B dataset and research infrastructure sold to AI and robotics customers, and the early revenue claims are notable but unevenly sourced. Multiple outlets reported 12 paying customers as of November 2026, while revenue claims above a $50 million annualized run rate rely largely on company-linked or single-source reporting and should be treated cautiously until corroborated by a broader set of independent disclosures [TechCrunch, November 2026] [TokenPost, November 2026] [Business Standard, November 2026].

Over the next 12 to 18 months, the key questions are straightforward: whether Humyn can convert early demand into durable repeatable contracts, whether field-data collection remains defensible as more Physical AI companies build internal pipelines, and whether its apparent U.S. customer concentration becomes an advantage or a dependency [Moneycontrol, November 2026] [TokenPost, November 2026]. The company has momentum, capital, and a timely wedge, but the public case still depends heavily on management-reported operating metrics rather than a long record of independently visible customer adoption.

Lightly corroborated -- Core company, founder, and financing facts are corroborated by TechCrunch, Bloomberg, and Business Standard; several operating metrics remain single-source or company-reported.

Taxonomy Snapshot

Axis Value
Stage Series A
Business Model B2B
Industry / Vertical Deeptech
Technology Type AI / Machine Learning
Geography Global / Remote-First
Growth Profile Venture Scale
Founding Team Co-Founders (2)
Funding $50M+ (total disclosed approximately $47,000,000)

How the Company Got Here

PUBLIC

Humyn Labs appears to have formed very recently, which matters because nearly every operating claim in market coverage sits on a short public record. The company is presented as a Bengaluru-headquartered platform focused on AI training data for robotics and Physical AI, with its website describing the business as supplying training data for robotics and related AI systems [Humyn Labs]. Crunchbase also lists Humyn Labs as a Bengaluru-based company and identifies Manish Agarwal and Ishank Gupta as founders [Crunchbase].

The public chronology is still sparse, but the available company-controlled sources point to a 2026 start. Crunchbase lists Humyn Labs as founded in 2026 [Crunchbase], and the company website positions the business around multimodal data capture and training data services for robotics and AI developers [Humyn Labs]. No state filing or legal entity record was provided in the source set, so the legal entity name and incorporation details are not established here.

The clearest public milestone is the company's emergence into broad coverage in November 2026, after which its profile became easier to verify across standard databases. Crunchbase reflects external financing activity for Humyn Labs [Crunchbase], while the company website shows an already expanded product surface spanning robotics data, voice data, image data, and transcription services under the same brand [Humyn Labs]. At this stage, the simplest view is that Humyn Labs is an early but unusually visible entrant in training data infrastructure for Physical AI, with headquarters and founding information corroborated publicly, but with limited filing-level detail in the materials reviewed.

Lightly corroborated -- Core company identity is corroborated by Crunchbase and the company website, but legal entity detail and filing-level milestones were not established from the provided sources.

Product and Technology

MIXED

The product is best understood as a data-creation system for Physical AI rather than a single annotation tool. Public reporting describes Humyn Labs as capturing people performing real-world tasks in homes, factories, warehouses, and farms, then converting those recordings into training datasets for robotics and world-model teams [The Times of India, November 2026] [TechCrunch, November 2026]. The company and third-party coverage both point to a multimodal scope: its website markets custom AI training data across voice, image, video, and sensor modalities, while TechCrunch reported that Humyn developed its own two-camera hardware to record everyday tasks and uses KGeN's contributor network in emerging markets to collect data inside homes [Humyn Labs] [TechCrunch, November 2026].

That combination matters because Humyn is selling more than raw footage. Its public materials describe collection, validation, processing, and labeling workflows, with provenance, consent documentation, and quality control positioned as part of the deliverable rather than an afterthought [Humyn Labs] [Business Standard, November 2026]. The website also advertises linguist-verified transcription, studio-quality voice data in 50-plus languages, image annotation, and multimodal dataset delivery, while secondary coverage adds synchronized touch, force, vision, depth, and 3D hand-pose data for teams training dexterous manipulation systems [Humyn Labs] [WhoTrainsAI]. The conservative read is that Humyn's differentiation rests on field operations and data acquisition infrastructure first, with proprietary hardware and QA processes as supporting layers, although most technical detail beyond the two-camera setup remains company-described rather than independently demonstrated [TechCrunch, November 2026] [Humyn Labs].

Lightly corroborated -- Core product scope is corroborated by TechCrunch and The Times of India, but many workflow and modality details rely on company website claims and secondary coverage.

Where the Demand Sits

PUBLIC

The market matters now because Humyn Labs sits in the supply chain for physical AI at the point where model ambition collides with a data bottleneck, and the public record around the company repeatedly frames demand in those terms rather than as generic annotation spend [TechCrunch, November 2026] [Bloomberg, November 2026].

Public sources do not provide a named third-party TAM, SAM, or SOM for Humyn Labs’ exact niche, which is multimodal, real-world human activity capture for robotics and world-model training. The safer read is to treat this as an overlap market spanning AI training data services, robotics development tooling, and physical AI evaluation, with Humyn positioned at the narrower end of that stack [TechCrunch, November 2026] [Moneycontrol, November 2026]. That narrowness matters. The company is not described as a broad enterprise data-labeling vendor in the reporting, but as a specialist provider to frontier AI labs, robotics teams building vision-language-action systems, and companies developing world models [Moneycontrol, November 2026] [TechCrunch, November 2026].

The demand drivers in the cited research are fairly consistent even when revenue claims vary in quality. TechCrunch, Bloomberg, and The Times of India all point to the same basic need: robotics and physical AI teams require recordings of people performing tasks in homes, factories, warehouses, and farms so models can learn manipulation and navigation in messy real environments [TechCrunch, November 2026] [Bloomberg, November 2026] [The Times of India, November 2026]. Humyn’s own reported operating footprint, more than 20 countries in Business Standard’s account, suggests that geography is not incidental but part of the product, since data gathered across household formats, work settings, and emerging-market environments may help customers avoid overfitting to a single setting or demographic [Business Standard, November 2026].

Adjacent and substitute markets are easier to identify than precise market size. One substitute is conventional AI data-labeling and synthetic-data workflows, which can cover image, video, and speech tasks but do not necessarily solve for egocentric capture, tactile context, or embodied task sequencing. Another adjacent market is robotics simulation and testing infrastructure, especially as The Economic Times reported that the company intends to move beyond collection toward testing robot performance in factories and warehouses, which would place it closer to deployment validation than pure data services [The Economic Times, November 2026]. A third adjacency is multilingual speech data, where Humyn’s website positions the company as a provider of transcription, voice collection, and multimodal training data across 50-plus languages, broadening the addressable spend beyond robotics if that offering is commercially material [Humyn Labs].

Macro and regulatory forces cut both ways. On the supportive side, the company’s customer base is reported to be heavily U.S.-oriented, and several articles tie demand to the recent capital formation around robotics and physical AI labs, which tends to increase outsourced spending on specialized datasets once model teams move from research demos to real-world performance targets [Moneycontrol, November 2026] [Bloomberg, November 2026]. On the constraining side, Humyn’s model depends on recording people in homes and workplaces across multiple jurisdictions, which raises consent, labor, and data-governance requirements even if the available public coverage does not spell out a regulatory issue specific to the company. That does not establish a current problem, but it does mean the market is not only a compute story, it is also an operations and compliance story.

Market slice What public sources support Evidence quality
Core niche Physical AI training data for robotics, frontier labs, and world-model developers YELLOW, multiple press reports describe the buyer set and use case [TechCrunch, November 2026] [Moneycontrol, November 2026]
Adjacent market Robotics testing and real-world deployment support YELLOW, reported strategic direction rather than disclosed revenue mix [The Economic Times, November 2026]
Adjacent market Multilingual speech and multimodal AI training data services RED, company website positioning without independent commercial validation [Humyn Labs]
Geographic demand concentration U.S.-led customer demand for a globally sourced data supply chain YELLOW, supported by reported U.S. revenue concentration and U.S. robotics customer framing [Moneycontrol, November 2026] [TechCrunch, November 2026]

The practical takeaway is that Humyn appears to be selling into a fast-forming procurement category rather than a mature line item with clean third-party market maps. That can be attractive if the category hardens around embodied AI workflows, but the absence of independent sizing data means the market case currently rests more on buyer urgency and category formation than on a validated TAM framework.

Lightly corroborated -- Market characterization is corroborated by TechCrunch, Bloomberg, Moneycontrol, The Times of India, and The Economic Times, but no independent named third-party market sizing report was provided for this exact category.

Competitive Landscape

Competitive set

MIXED Humyn Labs is positioning itself less as a generic labeling vendor and more as a field-data infrastructure layer for physical AI, which puts it in partial competition with specialist data suppliers, contributor networks, and in-house robotics data operations rather than only with annotation platforms [TechCrunch, November 2026] [The Times of India, November 2026] [Humyn Labs].

Company Positioning Stage / Funding Notable Differentiator Source
Humyn Labs Research and data platform for physical AI training datasets, focused on capturing human activity and turning it into training-ready data for robotics and world-model teams Series A, $47M disclosed in November 2026 Own two-camera hardware, uses KGeN contributor network, and sells multimodal collection across homes, factories, warehouses, and farms [TechCrunch, November 2026] [Business Standard, November 2026] [Humyn Labs]
AIChamp Named competitor in structured research Not established from cited public sources Not established from cited public sources [Structured facts]

The competitive map breaks into three layers. Second are adjacent data infrastructure providers, including broader AI training-data firms and contributor-network businesses, which can plausibly move into multimodal collection if customer demand consolidates around a smaller set of vendors; Humyn's own use of KGeN's network shows how close the boundary is between network access and productized data supply [TechCrunch, November 2026] [KGeN Foundation on X]. Third are internal teams at frontier AI labs and robotics companies, which remain the most credible substitute when data requirements are proprietary, safety-sensitive, or tightly coupled to model iteration cycles [TechCrunch, November 2026].

Humyn's edge today appears to come from orchestration rather than any one isolated asset. The company says it captures human activity across sound, sight, motion, and touch, and TechCrunch reported that it developed its own two-camera hardware while also using KGeN's emerging-market contributor network to collect data inside homes [Humyn Labs] [TechCrunch, November 2026]. That combination matters because physical-AI buyers often need collection, consent, field operations, and post-processing to work as one system. Still, the durability of that edge is not settled. Contributor networks can be shared, hardware can be copied if the design is not deeply proprietary, and service-heavy workflows tend to invite fast imitation once demand is proven. The part that could last longer is execution at geographic scale, if Humyn can keep quality consistent across programs in more than 20 countries as reported by Business Standard [Business Standard, November 2026].

The main exposure is that Humyn does not appear to own the customer relationship end-to-end in the way a platform with embedded tooling or an in-house lab does. KGeN is described as an input to Humyn's operating model, not merely a background supplier, which creates some dependency risk if access terms change or if similar networks are made available to rivals [TechCrunch, November 2026] [KGeN Foundation on X]. There is also a category boundary to watch: firms whose advantage is software embedded in robotics development stacks may have a stronger hold on iterative testing and deployment loops than a company centered on data capture. Public reporting hints that Humyn wants to move beyond raw collection toward real-world robot deployment support, but the evidence in the public record is still early and mostly company-framed through press coverage [The Economic Times, November 2026] [Moneycontrol, November 2026].

Over the next 18 months, the most plausible competitive outcome is a split market. Humyn is the likely winner if robotics labs continue outsourcing geographically distributed, multimodal data collection and value a vendor that can organize capture programs in homes and workplaces across multiple regions [The Times of India, November 2026] [Business Standard, November 2026]. The inverse scenario is also worth keeping on the table: Humyn would be more exposed if leading labs internalize collection and use vendors only for overflow capacity, because that would pressure differentiation down toward labor access and price rather than program design.

Single unverified source -- Core facts on Humyn are supported by TechCrunch, Business Standard, The Times of India, and the company website, but competitor-specific evidence is limited and the named rival in this section is not corroborated by independent public reporting in the materials provided.

Opportunity

PUBLIC

The prize here is unusually large if execution holds: Humyn Labs could become a default data infrastructure layer for physical AI teams that need real-world human activity data, not just synthetic simulations, at a moment when robotics and world-model builders are moving from demo systems toward deployment constraints [TechCrunch, November 2026] [Bloomberg, November 2026] [The Times of India, November 2026].

The headline opportunity is straightforward. If frontier AI labs and robotics companies increasingly discover that model progress is bottlenecked by scarce, consented, task-specific multimodal data from real environments, then a supplier that can repeatedly collect, validate, and package that data across geographies becomes more than a services vendor, it becomes operating infrastructure. Humyn Labs has some early evidence that this outcome is reachable rather than purely aspirational: the company says it has 12 paying customers as of November 2026, it has already raised a $47 million Series A composed of $26 million in equity led by Accel and $21 million in venture debt led by Lighthouse Canton Growth Debt Fund II, and multiple publications describe an operating model built around field capture in homes, factories, warehouses, and farms rather than only desk-based labeling [TechCrunch, November 2026] [Business Standard, November 2026] [The Times of India, November 2026]. That combination matters because the scarce asset in physical AI is not only annotation capacity, but the ability to organize repeatable real-world collection workflows at speed.

A second reason the upside is tangible is that Humyn appears to be aiming at a narrower, harder problem than generic data labeling. TechCrunch reported that the company developed its own two-camera hardware to record people performing everyday tasks and uses KGeN's contributor network to collect data in homes, while Bloomberg framed the end market as training robots to manipulate objects and operate in unpredictable settings [TechCrunch, November 2026] [Bloomberg, November 2026]. If that positioning holds, the company has a path to become a specialist provider for dexterous robotics, vision-language-action models, and world-model training workloads, categories where procurement may favor reliability and reproducibility over lowest-cost labor.

Scenario What happens Catalyst Why it's plausible
Physical AI system of record Humyn becomes the recurring data partner for frontier AI labs and robotics teams, expanding from custom capture into a standard layer for collection, validation, and labeling across programs A few early customers broaden spend from one-off datasets to ongoing pipelines as robotics teams move from research to deployment [TechCrunch, November 2026] [Bloomberg, November 2026] The company already reports 12 paying customers, and public coverage places it directly in the workflow of robotics and world-model builders rather than at the edge of the stack [TechCrunch, November 2026] [Moneycontrol, November 2026]
Emerging-market collection network at scale Humyn turns its access to distributed contributors and in-home capture into a defensible supply advantage for datasets that are hard to source in the U.S. or Europe alone KGeN-linked contributor access continues to lower collection friction across emerging markets, especially for home and workplace recordings [TechCrunch, November 2026] [KGeN Foundation 🟩 on X] Public reporting and company-adjacent sources repeatedly tie Humyn's model to KGeN's network and to operations spanning more than 20 countries, which suggests early coverage breadth even if the exact economics remain unverified [TechCrunch, November 2026] [Business Standard, November 2026]
Multimodal data vendor to deployment platform Humyn extends from data capture into evaluation and real-world testing support for robots in factories, warehouses, and similar settings Customers ask for downstream services once dataset procurement and model testing start to merge operationally [The Economic Times, November 2026] Retrieved reporting indicates management is already discussing a move beyond collection toward real-world deployment and performance testing, which is the natural adjacency if customers prefer fewer vendors in physical AI workflows [The Economic Times, November 2026]

The compounding logic is credible even from a thin public record. Each completed project can improve three assets at once: contributor access, collection playbooks, and task ontology. The more often Humyn captures human behavior across modalities and environments, the better it should become at quoting difficult projects, recruiting the right contributors, standardizing hardware and quality-control steps, and delivering formats that advanced model teams can use with less rework [Humyn Labs] [TechCrunch, November 2026]. The early signs of that flywheel are visible, though not yet deeply verified: the company markets multimodal capabilities across voice, image, video, and sensor data; TechCrunch reports proprietary capture hardware; and multiple outlets describe a customer base already paying for the output [Humyn Labs] [TechCrunch, November 2026] [TokenPost, November 2026]. If those inputs produce higher repeat purchase rates, Humyn could shift from project vendor to embedded research infrastructure.

The size of the win is easier to frame by role in the stack than by a confirmed market-size figure, because no reliable public TAM estimate was provided in the source set. A useful public comparable is not a direct peer but the value creation captured by infrastructure companies that become default suppliers to a technical bottleneck. Here, the most relevant benchmark is scenario-based: if Humyn were to become a leading independent provider of real-world physical AI datasets and adjacent testing infrastructure, with revenue quality closer to recurring infrastructure spend than to one-off annotation work, the outcome could support a multibillion-dollar valuation over time (scenario, not a forecast). That upside is not supported by disclosed valuation data, which remains absent from the public round coverage, but it is directionally grounded in the fact that Accel, Prosus Ventures, Jump Trading, and Lighthouse Canton backed the business at an unusually early stage, while management and press reports point to more than $50 million in annualized revenue run rate within months of launch, albeit on company-sourced claims that need caution [Bloomberg, November 2026] [Business Standard, November 2026] [Seeking Alpha] [TechCrunch, November 2026]. On a conservative reading, the best upside case is not that Humyn becomes a broad AI platform. It is that it becomes indispensable to the subset of AI companies that need costly, real-world, multimodal human activity data and cannot train around that constraint.

Single unverified source -- Section relies on a mix of independently reported funding and customer-count facts from TechCrunch, Bloomberg, Business Standard, and The Times of India, plus company-sourced product claims and less fully corroborated operating-scale and revenue assertions.

Sources

Public sources

  1. [TechCrunch, November 2026] India’s Humyn Labs raises $47M to feed U.S. robotics companies the data they need | https://techcrunch.com/2026/11/18/accel-backs-indias-humyn-labs-to-supply-training-data-needed-by-u-s-robotics-ai-developers/

  2. [Bloomberg, November 2026] Physical AI Startup Humyn Raises $47 Million From Accel, Prosus | https://www.bloomberg.com/news/articles/2026-11-18/physical-ai-startup-humyn-raises-47-million-from-accel-prosus

  3. [Business Standard, November 2026] AI startup Humyn Labs raises $47 million in Series A funding round | https://www.business-standard.com/companies/news/ai-startup-humyn-labs-raises-47-million-in-series-a-funding-round-126111800765_1.html

  4. [The Times of India, November 2026] Humyn Labs raises $47 mn for robot training | https://timesofindia.indiatimes.com/business/india-business/humyn-labs-raises-47-mn-for-robot-training/articleshow/135964647.cms

  5. [Forbes, March 2018] Nazara CEO On Being India’s Largest Gaming Company, Going Public, Growth Plans & More | https://www.forbesindia.com/article/startups/nazara-ceo-on-being-indias-largest-gaming-company-going-public-growth-plans-more/49827/1

  6. [Crunchbase] Humyn Labs - Crunchbase Company Profile & Funding | https://www.crunchbase.com/organization/humyn-labs

  7. [Humyn Labs] Humyn Labs | AI Training Data for Robotics & Physical AI | https://humynlabs.ai/

  8. [Moneycontrol, November 2026] Humyn Labs raises $47 million led by Accel to scale physical AI research and real-world deployment | https://www.moneycontrol.com/news/business/funding/humyn-labs-raises-47-million-led-by-accel-to-scale-physical-ai-research-and-real-world-deployment-14083680.html

  9. [WhoTrainsAI] Physical AI Data Companies | https://www.whotrains.ai/physical-ai-data-companies

  10. [The Economic Times, November 2026] Humyn Labs raises $47 million to scale physical AI, robotics deployment business | https://economictimes.com/tech/funding/humyn-labs-raises-47-million-to-scale-physical-ai-robotics-deployment-business/articleshow/135962272.cms

  11. [KGeN Foundation on X] KGeN Foundation post on X | https://x.com/KGeN_Official

  12. [TokenPost, November 2026] Humyn Labs raises $47M to accelerate physical AI training data expansion | https://www.tokenpost.com/news/business/19460

  13. [Bloomberg] Manish Agarwal, AXA Strategic Ventures US LLC: Profile and Biography | https://www.bloomberg.com/profile/person/17989471

Articles about Humyn Labs

View on Startuply.vc