MultiSet AI

Scan-agnostic visual positioning system (VPS) for Physical AI, converting reality-capture data into spatial maps.

Cover Block

From the public record

Field	Detail
Name	MultiSet AI
Tagline	Scan-agnostic visual positioning system (VPS) for Physical AI, converting reality-capture data into spatial maps [MultiSet AI, retrieved 2026]
Headquarters	San Francisco, USA [Crunchbase, retrieved 2026]
Founded	2024 [Crunchbase, retrieved 2026]
Stage	Seed [Crunchbase, retrieved 2026]
Business Model	API / Developer Platform [MultiSet AI, retrieved 2026]
Industry	Deeptech
Technology	AI / Machine Learning
Geography	Global / Remote-First [MultiSet AI, retrieved 2026]
Growth Profile	Venture Scale
Founding Team	Co-Founders (2): Nikhil Sawlani, Shadnam Khan [Prevu3D, November 2026]
Funding Label	Pre-seed [XROM, December 2025]
Total Disclosed	$250,000 [XROM, December 2025]

The Short Version

PUBLIC MultiSet AI is building a scan agnostic visual positioning system for physical AI, and it merits attention because the company is trying to turn reality capture data that enterprises already own into a localization layer for robots, smart glasses, phones, and field operations, a practical wedge in a market where capture workflow friction still slows adoption [MultiSet AI, retrieved 2026] [Prevu3D, November 2026]. Founded in 2024 in San Francisco, the company is still early, but the public record shows a product narrative that is unusually consistent across its website, partner materials, and developer surfaces, with emphasis on centimeter level localization, broad scan input support, and deployment flexibility across cloud, customer VPC, on premises, on device, and air gapped environments [MultiSet AI, retrieved 2026] [Prevu3D, November 2026].

The differentiation, according to company and partner materials, is less about inventing a new capture modality and more about making existing LiDAR point clouds, Matterport scans, 360 video, and phone capture usable as shared spatial ground truth, including a workflow where one capture can produce both machine localization output and a human readable Gaussian splat [MultiSet AI, retrieved 2026] [Prevu3D, November 2026]. That positioning is commercially sensible if it holds up in production, because it aligns with how industrial and robotics buyers already procure spatial data rather than asking them to restart from a proprietary stack [Prevu3D, November 2026].

The founding team is small and founder led. Public sources identify Nikhil Sawlani as co founder and CEO, with prior experience at Nextech AR Solutions according to XROM, while Shadnam Khan is identified as co founder and COO in partner materials and public profiles [XROM, December 2025] [Prevu3D, November 2026] [LinkedIn, retrieved 2026]. The team signal is relevant, though still only partially corroborated, because this category usually demands both deep computer vision execution and patient enterprise deployment work.

On capitalization, MultiSet has disclosed a $250,000 pre seed round in December 2025 with participation from a16z Scout Fund, Nivesha Ventures, and Bilawal Sidhu, while Crunchbase also lists $250,000 in total disclosed funding [XROM, December 2025] [Crunchbase, retrieved 2026]. The business model appears to be API or developer platform led, reinforced by the company pricing page, developer documentation, and two open roles spanning forward deployed deployment work and research science [MultiSet AI, retrieved 2026].

The next 12 to 18 months likely turn on proof that the company can translate product breadth into repeatable enterprise adoption. Public traction claims are promising but mixed in precision, ranging from 5,000 plus locations in 150 countries on the company site to 3,400 plus locations in 116 countries and 16 million plus square feet mapped in third party references, so investors should watch for more named deployments, clearer customer attribution, and evidence that the current Prevu3D partnership expands into durable production usage [MultiSet AI, retrieved 2026] [The Manila Times, June 2026] [AREA, retrieved 2026] [Prevu3D, November 2026].

Single-source, plausible -- This section uses a mix of company materials, a partner announcement, Crunchbase, and one funding report; core company, funding, and team facts are partially corroborated, while several traction and performance claims remain company-led or lightly verified.

Taxonomy Snapshot

Axis	Value
Stage	Seed
Business Model	API / Developer Platform
Industry / Vertical	Deeptech
Technology Type	AI / Machine Learning
Geography	Global / Remote-First
Growth Profile	Venture Scale
Founding Team	Co-Founders (2)
Funding	Pre-seed, total disclosed approximately $250,000 [XROM, December 2025]

The Company in Brief

PUBLIC MultiSet AI presents itself as infrastructure for physical-world localization, not as a vertical application layer. The company is based in San Francisco and was founded in 2024, according to its company profile and public database listing [MultiSet AI, retrieved 2026] [Crunchbase, retrieved 2026]. On its website, MultiSet describes the product in plain terms as a scan-agnostic visual positioning system for AR, robotics, and smart glasses, with positioning built from existing reality-capture inputs rather than a required proprietary capture workflow [MultiSet AI, retrieved 2026].

The public record is still thin, which is typical at this stage. Crunchbase identifies Nikhil Sawlani and Shadnam Khan as the founding team and classifies the business as a developer platform with disclosed pre-seed funding of $250,000 [Crunchbase, retrieved 2026]. MultiSet's website shows an already commercial-facing posture, with a developer portal, pricing page, documentation, product modules such as VPS, object tracking, and 3D reconstruction, and use-case positioning across robotics, oil and gas, pharmaceuticals, and construction [MultiSet AI, retrieved 2026].

A simple chronology emerges from those public sources. The company was founded in 2024 [Crunchbase, retrieved 2026], and by 2026 its website was advertising a production-oriented platform surface that included developer documentation, map versioning, and alternatives pages referencing Vuforia and Azure Spatial Anchors [MultiSet AI, retrieved 2026]. That does not by itself establish scale or customer quality, but it does suggest MultiSet reached the market with a defined product thesis early in its life.

Confirmed across multiple sources -- Confirmed by Crunchbase and the company website.

What They Have Built

Mixed sourcing

What stands out first is not a model claim but a workflow claim: MultiSet is trying to meet customers where their spatial data already exists. The company describes its product as a scan agnostic visual positioning system for physical AI that converts existing reality capture inputs into spatial maps for localization, with supported sources that include E57 and LiDAR point clouds, Matterport, 360 video, and phone capture, according to its site and a partner announcement [MultiSet AI, retrieved 2026] [Prevu3D, November 2026]. On its homepage, MultiSet says the system provides “human & machine readable maps,” and its visible product surfaces include VPS, object tracking, 3D reconstruction, map versioning, and conversion tools such as E57 to VPS, 3DGS to VPS, and 360 to VPS [MultiSet AI, retrieved 2026].

The technical pitch is that one capture can serve both machine navigation and human inspection workflows. MultiSet says a single capture can produce a machine localization map alongside a human readable Gaussian splat, and its materials describe centimeter level, 6 DoF localization for devices including robots, drones, smart glasses, phones, and wearables [MultiSet AI, retrieved 2026] [Augmented Enterprise Summit, retrieved 2026] [MultiSet AI, November 2025]. The same public materials emphasize deployment flexibility across cloud, customer VPC, on premises, on device, and air gapped environments, which matters more than usual in industrial settings where security and latency constraints often block cloud only products [MultiSet AI, retrieved 2026].

The evidence base is still early and leans heavily on company materials, but there are a few external signals worth separating from the marketing layer. Prevu3D's November 2026 partnership announcement independently supports the claim that MultiSet can ingest pre existing scan formats and is targeting industrial field team use cases [Prevu3D, November 2026]. Two current job postings, one for a forward deployed solutions engineer handling deployments from raw scan through production VPS and another for a senior research scientist focused on visual place recognition, localization, latency, and cost, also suggest the current stack is oriented around production mapping pipelines and computer vision research rather than a thin application wrapper (inferred from job postings) [MultiSet AI, retrieved 2026].

Unconfirmed -- Product scope is partly corroborated by Prevu3D and event materials, but several material capability claims rely on company website and company docs.

Market Size and Demand

PUBLIC

This market matters now because indoor and GPS-denied localization has moved from a demo-layer problem to an operating requirement for industrial AR, robotics, and spatial computing deployments that need repeatable positioning across real facilities [Prevu3D, November 2026] [MultiSet AI, retrieved 2026].

The evidence base for a formal TAM/SAM/SOM remains thin in the public record. No named third-party market report was captured for MultiSet AI’s exact category, so the cleaner approach is to frame the addressable market through adjacent public demand signals rather than force a synthetic market size. The company’s own materials place the product at the intersection of enterprise AR, connected worker software, robotics navigation, and digital-twin workflows, while the Prevu3D partnership points specifically to industrial field teams using existing spatial scans as an operational input rather than a one-time modeling asset [MultiSet AI, retrieved 2026] [Prevu3D, November 2026].

That positioning matters because the buyer problem is less about creating another 3D model and more about turning already-captured environments into a usable coordinate system for devices and workers. Public materials describe support for E57 and LiDAR point clouds, Matterport, 360 video, and phone capture, which suggests demand may be pulled by enterprises that already own fragmented scan data but lack a consistent localization layer across robots, smart glasses, phones, and drones [Prevu3D, November 2026] [MultiSet AI, retrieved 2026]. In practice, that shifts the relevant market from pure AR software into a broader infrastructure category spanning visual localization, spatial data interoperability, and industrial automation tooling [Prevu3D, November 2026].

The most visible tailwind in the cited research is reuse of existing reality-capture data. MultiSet’s stated wedge is scan agnosticism and flexible deployment across cloud, customer VPC, on-premises, on-device, and air-gapped settings, which lines up with the procurement needs of industrial and regulated environments where data residency, offline operation, or controlled infrastructure can shape product selection [MultiSet AI, retrieved 2026] [Prevu3D, November 2026]. The same source trail also points to a second tailwind: hardware proliferation. The company frames its maps as usable by robots, phones, smart glasses, headsets, drones, and facility workers, implying that value rises as more endpoint types require shared spatial ground truth rather than isolated app-specific anchors [MultiSet AI, retrieved 2026] [Augmented Enterprise Summit, retrieved 2026].

A third demand driver is the operationalization of industrial digital twins. Prevu3D’s announcement is notable not because it proves scale, it does not, but because it shows a named workflow where scan data collected for visualization can be extended into field execution and, eventually, autonomous systems [Prevu3D, November 2026]. If that pattern holds more broadly, the substitute market is not only competing VPS vendors but also manual asset-finding workflows, QR or fiducial-based indoor navigation, proprietary robotics mapping stacks, and digital-twin platforms that stop short of continuous localization [Prevu3D, November 2026] [MultiSet AI, retrieved 2026].

The adjacent markets are therefore easier to identify than the exact core market size. One adjacency is enterprise AR infrastructure, where persistent spatial registration is necessary for work instructions, remote assistance, and inspection overlays. Another is robotics and drone autonomy in GPS-denied settings, which the company explicitly cites as a use case. A third is spatial-commerce and planogram compliance, also listed in company materials, though public evidence here is limited to product positioning rather than disclosed deployments [MultiSet AI, retrieved 2026].

Market lens	What public evidence supports	Why it matters
Enterprise AR and connected worker	MultiSet and AREA materials tie the platform to enterprise AR and connected worker deployments [AREA, retrieved 2026] [MultiSet AI, retrieved 2026]	Suggests demand from workflows where persistent localization affects task completion and worker guidance
Industrial digital twin operations	Prevu3D partnership centers industrial field teams and existing scan data [Prevu3D, November 2026]	Points to a buyer set that already captures facilities and may want to extend scans into live operations
Robotics and drones in GPS-denied settings	Company use-case pages reference robots, drones, wearables, and assured PNT environments [MultiSet AI, retrieved 2026]	Expands the category beyond AR software into autonomy infrastructure
Spatial data interoperability	Supported inputs include LiDAR point clouds, Matterport, 360 video, and phone capture [Prevu3D, November 2026]	Implies value accrues where enterprises have heterogeneous legacy scan assets

The table points to a market that is best understood as converging demand from several existing budgets rather than a single clean software category. That can expand the opportunity if budgets are real, but it also means category definition and procurement ownership may stay unsettled in the near term.

On regulation and macro forces, the public sources support a narrow but relevant point: deployment flexibility matters because some customers need private cloud, on-premises, on-device, or air-gapped operation, especially where facility data, inspection workflows, or autonomy systems cannot sit in a standard multi-tenant cloud stack [MultiSet AI, retrieved 2026] [Prevu3D, November 2026]. Beyond that, the source set does not justify broad claims about specific regulatory catalysts. The more defensible macro read is that enterprises are already spending on reality capture and industrial digitization, and vendors that can repurpose those sunk data assets into recurring operational infrastructure may benefit from tighter ROI scrutiny than companies that require greenfield capture programs [Prevu3D, November 2026] [MultiSet AI, retrieved 2026].

Single-source, plausible -- This section relies on one independent partner announcement and trade association materials, with category framing partly supported by company materials and no named third-party market size report in the captured sources.

Who Else Is Fighting for This

MIXED MultiSet AI is positioning itself less as another capture stack and more as an interoperability layer for spatial localization, which puts it in direct comparison with established VPS platforms like Vuforia and Azure Spatial Anchors, while also competing indirectly with digital-twin and AR workflow vendors that already sit on enterprise scan data [MultiSet AI, retrieved 2026] [Prevu3D, November 2026].

Company	Positioning	Stage / Funding	Notable Differentiator	Source
MultiSet AI	Scan-agnostic VPS for Physical AI, turning existing reality-capture data into spatial maps for localization across robots, phones, smart glasses, and other devices	Founded 2024; disclosed pre-seed of $250,000 in December 2025	Accepts customer-owned scan formats and supports cloud, customer VPC, on-premises, on-device, and air-gapped deployments	[MultiSet AI, retrieved 2026] [Prevu3D, November 2026] [XROM, December 2025]
Vuforia	Enterprise AR platform with alternatives coverage referenced by MultiSet on its site	Not established in the provided public source set	Recognized incumbent in industrial AR workflows, with relevance as a point of comparison in MultiSet's own alternatives navigation	[MultiSet AI, retrieved 2026]
Azure Spatial Anchors	Cloud-based spatial anchoring product referenced by MultiSet on its site	Not established in the provided public source set	Strong adjacency to Microsoft's enterprise and developer distribution, especially where teams already build on Azure	[MultiSet AI, retrieved 2026]

The competitive map breaks into three groups. Incumbents include Vuforia, which is already associated with enterprise AR deployment paths, and Azure Spatial Anchors, which benefits from Microsoft's broader developer and cloud footprint even where the product itself is only one layer in a larger stack [MultiSet AI, retrieved 2026]. Challengers are thinner on the public record in this dataset, but MultiSet's own pitch and the Prevu3D partnership suggest it is trying to wedge into accounts that already possess LiDAR, Matterport, E57, 360 video, or phone-captured scans, rather than forcing a new capture standard from day one [Prevu3D, November 2026]. Adjacent substitutes matter as well: digital-twin vendors, industrial AR workflow software, and robotics localization stacks can absorb some of the same budget even if they do not present as direct VPS competitors in branding terms [Prevu3D, November 2026] [MultiSet AI, retrieved 2026].

The most credible edge visible in public materials is deployment flexibility tied to scan agnosticism. MultiSet says customers can bring existing scan data and deploy in cloud, customer VPC, on-premises, on-device, or air-gapped environments, a combination that maps cleanly to regulated and operationally sensitive industrial settings where buyers often resist net-new hardware or centralized cloud dependence [MultiSet AI, retrieved 2026] [Prevu3D, November 2026]. That edge looks real if the technical claims hold, but it is also perishable. An interoperability advantage is durable only if it compounds into data, integrations, and production references faster than larger platforms can replicate feature parity.

The exposure is equally clear. Azure Spatial Anchors has an obvious distribution advantage wherever buyers are already standardized on Microsoft tooling, and Vuforia has longer-standing mindshare in enterprise AR use cases that overlap with field operations and industrial workflows [MultiSet AI, retrieved 2026]. MultiSet's disclosed funding is still modest at $250,000 pre-seed, according to XROM and Crunchbase, which likely limits how quickly it can build enterprise sales coverage, solution engineering, and ecosystem partnerships relative to larger incumbents [XROM, December 2025] [Crunchbase]. The company also does not appear, from the public source set, to own a dominant channel such as a cloud platform, device OS, or installed hardware base.

The most plausible 18-month scenario is a split market rather than a single-takeout outcome. Winner if interoperability becomes the buyer's first filter is MultiSet, particularly in brownfield industrial environments where enterprises already have spatial data and need a VPS layer that can run in private or constrained deployments [Prevu3D, November 2026] [MultiSet AI, retrieved 2026]. Loser if cloud-native bundling becomes decisive is standalone VPS vendors generally, with MultiSet especially exposed if Microsoft can make Azure Spatial Anchors functionally good enough inside a broader enterprise contract motion [MultiSet AI, retrieved 2026]. By the same logic, Vuforia remains positioned if AR workflow ownership matters more than localization infrastructure, though the current source set does not establish its latest product or funding posture.

Single-source, plausible -- Core positioning and named competitor references are confirmed by MultiSet's public site and the Prevu3D partner announcement, but competitor funding, current product scope, and comparative performance are not independently established in the provided source set.

Opportunity

From the public record The prize here is not a feature sale, it is the chance to become a horizontal localization layer for indoor and GPS-denied physical environments, a role that would sit underneath robots, wearables, phones, and industrial software rather than compete with any one of them [MultiSet AI, retrieved 2026] [Prevu3D, November 2026].

The headline opportunity is straightforward. If MultiSet can reliably turn scans customers already own into a shared spatial map, then it can lower one of the harder adoption barriers in industrial AR and robotics, namely the cost and friction of creating localization-ready environments from scratch [Prevu3D, November 2026] [MultiSet AI, retrieved 2026]. That outcome is still early, but it is not purely aspirational: the company has a disclosed pre-seed syndicate, named as a16z Scout Fund, Nivesha Ventures, and Bilawal Sidhu, a live partner announcement with Prevu3D, and public claims of production-scale usage that extend beyond a single pilot narrative, even if some of those operating metrics remain company- or trade-source based [XROM, December 2025] [Crunchbase, retrieved 2026] [Prevu3D, November 2026] [The Manila Times, June 2026].

The upside paths are easiest to see when broken into concrete adoption motions rather than one omnibus enterprise story.

Scenario	What happens	Catalyst	Why it's plausible
Industrial systems layer	MultiSet becomes the localization back end inside industrial digital-twin, field-service, and connected-worker workflows	The Prevu3D partnership turns into repeat integrations for industrial field teams using existing scans	Prevu3D has already announced the partnership and described MultiSet's role in making spatial intelligence usable for field workers and later autonomous systems [Prevu3D, November 2026]
Embedded developer API	MultiSet wins developers that need scan-agnostic VPS without building their own capture and mapping stack	Ongoing SDK expansion and install-free AR support widen the addressable developer base	MultiSet says Version 1.6.5 added expanded SDK support and install-free AR capabilities, while its docs and pricing pages position the product as a developer platform [MultiSet AI, retrieved 2026] [MultiSet AI, retrieved 2026]
Physical AI infrastructure	MultiSet becomes a common coordinate layer across robots, drones, smart glasses, and mobile devices in GPS-denied sites	A category shift toward mixed fleets that need one shared map across humans and machines	The company explicitly positions its VPS as one map that both machines and humans can read, and partner materials describe target devices that span robots, phones, and wearables rather than a single endpoint [MultiSet AI, retrieved 2026] [Prevu3D, November 2026]

The compounding dynamic would come from reuse. Every time a customer can ingest spatial data it already owns, deploy in its preferred environment, and use the same map across multiple device types, the switching cost rises and the sales motion broadens from one application to a site-wide or fleet-wide standard [Prevu3D, November 2026] [MultiSet AI, retrieved 2026]. Public evidence hints that this may already be the intended flywheel: MultiSet markets support for cloud, customer VPC, on-premises, on-device, and air-gapped deployments, and it claims scale across thousands of locations and hundreds of countries, which, if borne out, would suggest the product is being pitched as infrastructure rather than a bespoke services layer [MultiSet AI, retrieved 2026] [The Manila Times, June 2026].

The size of the win is necessarily scenario-based because there is no confirmed market-sizing dataset in the source pack. A reasonable public comparable frame is not current revenue, which is undisclosed, but strategic role: if MultiSet were to become a default VPS and spatial-ground-truth API for industrial environments, the outcome would resemble infrastructure software attached to robotics, AR, and digital-twin budgets rather than a narrow app vendor [MultiSet AI, retrieved 2026] [Prevu3D, November 2026]. On that basis, a credible upside case is a venture-scale platform outcome measured in the hundreds of millions of dollars of enterprise value, and potentially more if the company becomes embedded across multiple device ecosystems and industrial software partners (scenario, not a forecast). The evidence does not yet support a tighter valuation range, but the combination of scan-agnostic ingestion, flexible deployment, early partner validation, and reported global footprint is enough to keep the upside case open [XROM, December 2025] [Prevu3D, November 2026] [The Manila Times, June 2026].

Single-source, plausible -- Relies on one independent news item, one partner announcement, Crunchbase, and several company claims that are only partially corroborated.

Sources