Mosaico
Open-source data engine for robotics sensor data and physical AI.
Website: https://mosaico.dev/
Cover Block
From the public record
| Field | Detail |
|---|---|
| Name | Mosaico |
| Tagline | Open-source data engine for robotics sensor data and physical AI. [mosaico.dev, retrieved 2026] [Startups Magazine, November 2026] |
| Headquarters | Parma, Italy [Startups Magazine, November 2026] |
| Founded | 2023 [Startups Magazine, November 2026] |
| Stage | Seed [Startups Magazine, November 2026] |
| Business model | Open Source / Commercial |
| Industry | Deeptech |
| Technology | Robotics |
| Geography | Western Europe |
| Growth profile | Venture Scale |
| Founding team | Gabriele Lini, Federico Cabassi, Francesco Di Corato [Startups Magazine, November 2026] |
| Funding label | Seed |
| Total disclosed funding | $8,070,000 [Startups Magazine, November 2026] |
Links
From the public record
- Website: https://mosaico.dev/
- Careers: https://mosaico.dev/careers/
- Docs: https://docs.mosaico.dev/
- GitHub: https://get.mosaico.dev/
The Short Version
PUBLIC Mosaico is building open-source data infrastructure for robotics sensor data, and it merits attention now because the company paired a fresh €7.5 million seed round with a clear pitch into the bottleneck between robot data collection and physical AI model training [Startups Magazine, November 2026] [Startup Researcher, November 2026]. Founded in 2023 in Parma, Mosaico is positioning itself around a problem robotics teams know well: sensor data arrives in large volumes and awkward formats, and too much engineering time disappears into ingestion, transformation, storage, and query plumbing rather than model or application work [Startups Magazine, November 2026] [Headstart, November 2026].
The product case is straightforward on paper: Mosaico describes the platform as an open-source, pluggable data engine for robotics sensor data, while third-party coverage says it supports both plain-language and SQL-style queries, a useful combination if the company can make heterogeneous robotics datasets meaningfully searchable without forcing teams into bespoke pipelines [Startups Magazine, November 2026] [Startup Researcher, November 2026]. Company documentation adds technical color, including support for ROS bags, MCAP, and custom sensors, plus an on-premise deployment model, but those details remain primarily company-sourced and should be treated as product claims rather than independently verified market adoption signals [Mosaico, retrieved 2026] [Mosaico Doc, retrieved 2026].
The founding team is the strongest public proof point at this stage. CEO Gabriele Lini, CTO Federico Cabassi, and CSO Francesco Di Corato are identified as former Ambarella engineers, and multiple sources describe them as having spent roughly 12 to 15 years building autonomous-vehicle systems and data pipelines between Parma and Silicon Valley [Startups Magazine, November 2026] [The Robotics Media, November 2026] [Medium, retrieved 2026].
On financing, Mosaico disclosed a seed round of €7.5 07 million, but the euro amount is the directly reported figure and the cleaner one to anchor on here [Startups Magazine, November 2026] [DevCuration, November 2026]. The commercial model appears to be open source with a commercial layer, which is plausible for infrastructure sold into technical users, though public evidence does not yet establish customer traction, deployment scale, or monetization maturity beyond the platform narrative itself [Startups Magazine, November 2026] [Mosaico, retrieved 2026].
Over the next 12 to 18 months, the practical questions are less about category relevance and more about execution: whether Mosaico can convert technical credibility into named users, whether its open-source distribution creates a repeatable commercial motion, and whether expansion into the United States produces customer evidence rather than just footprint ambition [DevCuration, November 2026] [Headstart, November 2026]. The setup is promising, but the public record still reads like an early infrastructure company with a strong founding fit and a timely thesis, not yet one with enough independent operating data to underwrite adoption velocity confidently [Startups Magazine, November 2026] [Startup Researcher, November 2026].
Single-source, plausible -- Core company, founder, and funding facts are corroborated by multiple public sources, while several product details rely on company materials.
Taxonomy Snapshot
| Axis | Value |
|---|---|
| Stage | Seed |
| Business Model | Open Source / Commercial |
| Industry / Vertical | Deeptech |
| Technology Type | Robotics |
| Geography | Western Europe |
| Growth Profile | Venture Scale |
| Founding Team | Co-Founders (3+) |
| Funding | Seed, total disclosed approximately $8,070,000 [Startups Magazine, November 2026] |
The Company in Brief
PUBLIC
Mosaico enters the record first as a product-led infrastructure company rather than a heavily narrated origin story. The company presents itself as "the data engine for physical AI" and describes its platform as open-source infrastructure for robotic sensor data on its website [mosaico.dev, retrieved 2026]. Structured company facts place headquarters in Parma, Italy and founding in 2023 [mosaico.dev, retrieved 2026].
The public timeline is still short, but the sequence that matters is clear. By 2026, Mosaico had put forward a public web presence, product documentation, and a careers page under the mosaico.dev domain, indicating an operating software organization built around robotics data infrastructure [mosaico.dev, retrieved 2026] [Mosaico Doc, retrieved 2026] [Mosaico, retrieved 2026]. In November 2026, the company was reported to have raised a €7.5 million seed round led by United Ventures, with participation from OpenOcean and Connect Ventures, a milestone that moved it from early product formation into externally financed expansion [Startups Magazine, November 2026].
A few boundaries are worth keeping explicit. The available company-source material supports the product identity and web footprint, but it does not establish a publicly cited legal entity name in this section's permitted source set [mosaico.dev, retrieved 2026]. It also helps to separate Mosaico, the Italian robotics data startup, from the unrelated finance software company Mosaic, which appears frequently in search results under a near-identical name [mosaico.dev, retrieved 2026].
Single-source, plausible -- Company website confirms product identity and operating footprint; founding year and headquarters rely on structured company facts tied to the company source set, while the November 2026 financing milestone is supported by Startups Magazine rather than the preferred source set for full independent corroboration.
What They Have Built
MIXED
Mosaico is positioning itself at a fairly specific infrastructure layer: not robot control software, and not a model company, but the data plumbing that sits between sensor exhaust and training or analysis workflows. Public reporting describes the product as an open-source, pluggable data engine for robotics sensor data, built to ingest, transform, store, and query large volumes of information produced by robot sensors [Startups Magazine, November 2026]. Startup Researcher adds that the platform supports both plain-language and SQL-style queries, which, if borne out in production use, would make the product legible to both robotics engineers and more conventional data teams [Startup Researcher, November 2026].
The company website and documentation provide a more detailed, but company-sourced, picture of how that stack is meant to work. Mosaico says the platform runs on a customer's own infrastructure, including servers, cloud accounts, or edge hardware [PUBLIC] [Mosaico, retrieved 2026]. Its docs say the system ingests ROS bags, MCAP, and custom sensor inputs through a structured interface with automatic schema translation, and that the Python SDK is organized around three primitives, push, find, and stream [PUBLIC] [Mosaico, retrieved 2026] [Mosaico Doc, retrieved 2026]. The same documentation describes a Rust daemon called mosaicod communicating over Apache Arrow, ontology-based querying across heterogeneous data sources, and a Watchdog feature for automatic ingestion of MCAP and ROS bag files [PUBLIC] [Mosaico Doc, retrieved 2026]. Those details are useful for understanding product intent, but they remain company assertions rather than independently verified deployment evidence.
One recent product surface appears to be Mosaico Alchemy, which the homepage presents as a new release and Marco Rizza describes as an open-source Python toolkit with a robotic manipulation pack for importing public robotics datasets [Mosaico, retrieved 2026] [Marco Rizza]. Taken together, the public material suggests an effort to standardize robotics data handling from ingestion through retrieval, with an emphasis on making fragmented sensor data queryable and reusable across training workflows [Startups Magazine, November 2026] [Mosaico Doc, retrieved 2026]. The technical ambition is clear; what remains less clear from public evidence is how much of this stack is already hardened in production environments beyond demos, docs, and launch materials.
Unconfirmed -- This section combines independent press coverage with material product detail that is sourced primarily from Mosaico's own website and documentation, without independent verification of real-world deployment.
Market Size and Demand
PUBLIC
The market matters now because physical AI has moved from model experimentation to data bottlenecks, and Mosaico is positioned at the layer where robotics teams have to make sensor data usable before they can train, test, or debug anything downstream [Startups Magazine, November 2026] [Startup Researcher, November 2026].
The hard part in sizing this market is that the company is early and the available source set does not include a named third-party TAM study specific to robotics sensor-data infrastructure. Public reporting does, however, frame the demand problem consistently. Startups Magazine describes Mosaico as an open-source, pluggable data engine for robotics sensor data, built to ingest, transform, store, and query large volumes of machine-generated data [Startups Magazine, November 2026]. Startup Researcher adds that the platform supports plain-language and SQL-style queries, which places it at the intersection of data infrastructure and applied robotics tooling rather than at the model layer itself [Startup Researcher, November 2026]. On that evidence, the addressable market is best understood as an adjacent slice of the broader robotics software stack and the data-infrastructure spend that supports autonomy teams.
Because no source here provides a direct TAM, the cleaner approach is to define the market through the workload. Mosaico is selling into teams that collect high-volume multimodal sensor data, need to normalize formats, and want consistent retrieval across training and development workflows [Mosaico, retrieved 2026] [Mosaico Doc, retrieved 2026]. That suggests a serviceable market made up of robotics developers, autonomous systems teams, and physical-AI researchers who have already outgrown ad hoc internal pipelines. The narrower obtainable market, at least in the near term, is likely the subset willing to adopt open-source infrastructure with a commercial layer, but the public record does not yet show customer counts, deployments, or pricing that would support a quantified SOM [Headstart, November 2026].
The demand drivers in the source set are straightforward. First, the founders' background points to a persistent operational pain point: multiple reports say the team spent roughly 12 to 15 years building autonomous-vehicle systems and data pipelines, including time at Ambarella, and identified repeated friction around DIY data infrastructure and debugging overhead [The Robotics Media, November 2026] [Startups Magazine, November 2026] [Connect Ventures Opinions]. Second, the product architecture is aimed at scale and deployment control. The company says the platform runs on a user's own servers, cloud accounts, or edge hardware, and its documentation highlights support for ROS bags, MCAP, and custom sensors through a structured interface with schema translation [Mosaico, retrieved 2026] [Mosaico Doc, retrieved 2026]. Those features line up with a market where data locality, format fragmentation, and reproducibility matter as much as raw storage.
Adjacent and substitute markets are easier to identify than direct market size. One adjacency is general-purpose data infrastructure, where teams might otherwise assemble object storage, metadata services, ETL tooling, and analytics layers themselves. Another is MLOps and dataset-management software for AI teams, which addresses parts of the workflow but may not be optimized for robotics-specific sensor formats or ontology-driven querying [Mosaico Doc, retrieved 2026]. A third substitute is the in-house pipeline, which is the problem the company itself is explicitly targeting through its open-source positioning and developer tooling [dilica.it] [Mosaico, retrieved 2026].
Macro and regulatory forces in the public evidence are suggestive rather than conclusive. The strongest macro tailwind is simple: more robotics development produces more sensor data, and more model-centric autonomy work raises the cost of poor data retrieval and preparation [Startups Magazine, November 2026] [Headstart, November 2026]. There is also a deployment-control angle. Mosaico's on-premise positioning may appeal in environments where teams want tighter control over data handling or compute placement, but the current source set does not establish specific regulatory certifications, compliance claims, or sector-specific mandates, so that inference should be kept narrow [Mosaico, retrieved 2026].
| Market lens | What the sources support | Evidence quality |
|---|---|---|
| Core market | Robotics sensor-data infrastructure and data plumbing for physical AI teams | Multiple public reports [Startups Magazine, November 2026] [Startup Researcher, November 2026] |
| Adjacent market | Broader data infrastructure for high-volume machine data workflows | Product and documentation support, partly company-sourced [Mosaico, retrieved 2026] [Mosaico Doc, retrieved 2026] |
| Substitute market | In-house pipelines and general MLOps or analytics stacks | Inference from company positioning and founder problem statement [Connect Ventures Opinions] [dilica.it] |
| Quantified TAM/SAM/SOM | No named third-party sizing study in the current public source set | Limited |
The table shows why the market case is directionally credible but still light on third-party quantification. The need is well described; the size of the spend pool, by contrast, still has to be established through customer adoption and category formation rather than market-report arithmetic.
Unconfirmed -- Core market framing is corroborated by multiple public reports, but the section's material sizing premises rely on company and investor descriptions without an independent third-party TAM study.
Who Else Is Fighting for This
MIXED The competitive picture is still defined less by named rivals than by the shape of the problem: Mosaico is positioning itself between internal robotics data tooling, general-purpose data infrastructure, and emerging physical-AI software stacks, with an open-source wedge aimed at teams that treat sensor-data management as a bottleneck rather than a side task [Startups Magazine, November 2026] [Mosaico, retrieved 2026].
The first segment is the status quo, internal pipelines built by robotics teams themselves. That framing comes through repeatedly in public materials: Mosaico is described as handling the "data plumbing" of ingesting, transforming, storing, and querying robot-sensor data, and the founders' prior experience is presented as years spent dealing with DIY pipelines in autonomy systems [Startups Magazine, November 2026] [Connect Ventures Opinions] [The Robotics Media, November 2026]. In practice, that means Mosaico is not only competing with startups. It is also competing with the decision by a well-funded robotics company to keep using ROS bags, custom storage layers, and ad hoc query workflows assembled in-house, especially where engineering teams already view data infrastructure as core intellectual property.
The second segment is adjacent infrastructure, meaning general data engines and developer tools that were not built specifically for robotics but can absorb portions of the workload. Public evidence does not name direct competitors, so any tighter market map would be speculative. What can be said on the record is that Mosaico is trying to differentiate on robotics-specific ergonomics: support for ROS bags, MCAP, custom sensors, ontology-based querying across sources, and a Python SDK built around push, find, and stream rather than a conventional analytics workflow [Mosaico Doc, retrieved 2026] [Startup Researcher, November 2026]. That is a real product distinction, but it is also a perishable one if broader data-platform vendors or open-source projects decide the robotics and physical-AI workflow is large enough to justify dedicated adapters and schema layers.
Where Mosaico appears strongest today is talent-market fit and product framing. All three co-founders are reported former Ambarella engineers, and coverage ties them to roughly 12 to 15 years of work on autonomous-vehicle systems and data pipelines, which matters because this category rewards teams that have seen the failure modes firsthand rather than abstractly [Startups Magazine, November 2026] [The Robotics Media, November 2026] [Medium]. The open-source posture may also help with developer adoption and trust in infrastructure decisions, especially where robotics teams want on-premise deployment and control over data location [Mosaico, retrieved 2026] [dilica.it, retrieved 2026]. That edge is meaningful, though not obviously durable by itself. Founder credibility can open doors early, but long-term defensibility usually has to come from ecosystem adoption, workflow lock-in, or a dataset and ontology standard that others begin to build around. Public evidence does not yet establish that level of network effect.
The exposure is equally clear. Mosaico has a specific thesis around robotics sensor data, but no public source in the record names customers, production deployments, or commercial partners, which makes it hard to judge whether the company is already becoming infrastructure of record for any part of the market [Headstart, November 2026] [Startups Magazine, November 2026]. That leaves an opening on two fronts. Large robotics programs can continue building internally, while adjacent infrastructure vendors can move downward from broader data tooling if customer demand sharpens. The company also does not appear to own a distribution channel beyond product-led open-source adoption and founder networks, at least from the public record, and that can be a slower route in enterprise robotics than categories where self-serve usage naturally converts to paid expansion [Mosaico, retrieved 2026] [Startups Magazine, November 2026].
The most plausible 18-month scenario is that the competitive "winner if X" is Mosaico itself, if open-source usage turns into a de facto workflow standard for handling ROS and MCAP-heavy sensor datasets across research and commercial robotics teams [Mosaico Doc, retrieved 2026] [Startups Magazine, November 2026]. The corresponding "loser if Y" is the in-house toolchain, if robotics teams conclude that maintaining bespoke ingestion, ontology, and query infrastructure is too costly relative to buying into an opinionated platform from a team with domain depth [Connect Ventures Opinions] [The Robotics Media, November 2026]. The reverse scenario is also plausible: if adoption stalls before community and workflow lock-in emerge, internal stacks remain the default incumbent, and Mosaico risks being treated as a useful tool rather than a system of record.
Inferred, not confirmed -- Core positioning is corroborated by public press coverage and company documentation, but the competitive map lacks named direct competitors in the source set and therefore relies partly on category-level inference from public product evidence.
Opportunity
PUBLIC
The prize is unusually large if Mosaico executes, because the company is trying to become the default data layer for robotics teams that already drown in sensor output but still stitch together storage, ingestion, and query infrastructure by hand [Startups Magazine, November 2026] [The Robotics Media, November 2026].
The headline opportunity is straightforward: Mosaico could become a category-defining infrastructure platform for robotics and physical AI, not by owning the model layer but by owning the data plumbing beneath it. That reading is supported by the product wedge already described in public sources, namely an open-source, pluggable data engine built to ingest, transform, store, and query robotics sensor data, with support for plain-language and SQL-style queries [Startups Magazine, November 2026] [Startup Researcher, November 2026]. It is also supported by the team fit. Public reporting ties the three co-founders, Gabriele Lini, Federico Cabassi, and Francesco Di Corato, to Ambarella and to roughly 12 to 15 years building autonomous-vehicle systems and data pipelines, which narrows the usual concern that a young infrastructure company is solving an abstract problem rather than one its founders have handled directly [Startups Magazine, November 2026] [The Robotics Media, November 2026].
The evidence remains early, but the path is reachable rather than purely aspirational because the company has already raised a €7.5 million seed round led by United Ventures with participation from OpenOcean and Connect Ventures, and the stated use of proceeds includes product development, global open-source distribution, and expansion into the United States [Startups Magazine, November 2026] [DevCuration, November 2026]. In infrastructure markets, that combination matters: a technical wedge, open-source distribution, and fresh capital behind go-to-market can create a real shot at becoming the system of record before larger incumbents focus on the niche.
| Scenario | What happens | Catalyst | Why it's plausible |
|---|---|---|---|
| Open-source standard layer for robotics data | Mosaico becomes the common ingestion and query layer used by robotics developers across research and commercial environments | Wider adoption of the open-source engine and developer tooling such as the Python SDK and documentation hub [Mosaico, retrieved 2026] [Mosaico Doc, retrieved 2026] | Public materials consistently position the product as open-source infrastructure for robotics sensor data, and the company says data management should be foundational rather than premium in robotics development [Startups Magazine, November 2026] [dilica.it, retrieved 2026] |
| Physical AI data backbone for enterprise robotics teams | Mosaico moves from developer tool to production data backbone for teams managing petabyte-scale sensor pipelines | Expansion into the United States after the seed round, plus product releases such as Mosaico Alchemy [DevCuration, November 2026] [Mosaico, retrieved 2026] | The homepage and docs frame the system around petabyte-scale sensor handling, on-premise deployment, ontology, ingestion, and query, which fits enterprise-grade robotics workflows if execution holds [Mosaico, retrieved 2026] [Mosaico Doc, retrieved 2026] |
| Embedded data substrate for simulation and training workflows | Mosaico becomes the default bridge between raw robot logs and training-ready assets for physical AI | Continued tooling around dataset preparation and simulation workflows, including Alchemy and MuJoCo-linked examples [Marco Rizza, retrieved 2026] [Mosaico Doc, retrieved 2026] | Public docs already show replaying queried robot joint-state data in MuJoCo, while the company describes the platform as turning robotic sensor data into training-ready assets [Mosaico Doc, retrieved 2026] [Mosaico, retrieved 2026] |
What compounding looks like here is less a classic consumer network effect and more an infrastructure flywheel. If a robotics team adopts Mosaico first for ingestion, then standardizes schemas and ontology across ROS bags, MCAP, and custom sensors, the switching cost rises because downstream search, retrieval, debugging, and simulation workflows begin to depend on that normalized layer [Mosaico, retrieved 2026] [Mosaico Doc, retrieved 2026]. Each additional connector, ontology definition, and workflow primitive such as push, find, and stream can make the next team deployment easier, especially if open-source adoption broadens the surface area before commercial monetization is fully visible [Mosaico, retrieved 2026].
There is also an early distribution logic that could compound if the open-source thesis holds. A company that lets developers start on their own infrastructure, query across different data sources once they share the same ontology, and automate ingestion of ROS bag and MCAP files has a chance to spread bottom-up inside robotics organizations before a formal platform purchase happens [Mosaico Doc, retrieved 2026]. That does not prove commercial conversion, but it is the right shape of motion for infrastructure software, and it is one reason the upside case deserves attention despite the limited public traction record.
The size of the win is hard to quantify from public evidence because no verified market-size study or close public pure-play comparable appears in the source set. Even so, there is one external marker worth using cautiously: the separate finance software company Mosaic, which shares a similar name but is unrelated, raised $73 million in disclosed funding by June 2023 as a category software platform in its own domain [TechCrunch, June 2023]. That is not a direct comparable on product or market, but it illustrates how infrastructure software with cross-functional importance can attract substantial enterprise value once it becomes system-critical. For Mosaico, the more relevant framing is qualitative: if the company becomes the default data substrate for robotics and physical AI teams, it could plausibly support a venture-scale outcome in the hundreds of millions of dollars of enterprise value (scenario, not a forecast), with the inflection driven by standard-setting inside a young but expanding software layer of robotics rather than by hardware economics alone [Startups Magazine, November 2026] [Mosaico, retrieved 2026].
Inferred, not confirmed -- Core upside framing is grounded in public funding coverage and company documentation, but traction, market size, and compounding assumptions rely partly on company-described product capabilities without independent customer validation.
Sources
From the public record
[mosaico.dev, retrieved 2026] Mosaico | Next-Gen Robotics | https://mosaico.dev/
[Startups Magazine, November 2026] Mosaico raises €7.5M in seed funding round | https://startupsmagazine.co.uk/mosaico-raises-e7-5m-in-seed-funding-round
[Startup Researcher, November 2026] Mosaico Raises €7.5 Million Seed Round for Robotics Data | https://www.startupresearcher.com/news/mosaico-raises-7-5-million-seed-round-for-robotics-data
[Headstart, November 2026] Exclusive: Italian robotics startup Mosaico raises €7.5m to tackle AI’s data problem | https://www.headstart.news/articles/f02b674a4393bfcc5a6e77ae0d7b2db4
[Mosaico, retrieved 2026] Join the future of Physical AI | Mosaico | https://mosaico.dev/careers/
[Mosaico Doc, retrieved 2026] Overview | Mosaico Doc | https://docs.mosaico.dev/overview/
[The Robotics Media, November 2026] Mosaico Raises $8.3M Seed for Robotics Sensor Data Platform | https://theroboticsmedia.com/article/mosaico-raises-8-3m-seed-robotics-sensor-data
[Medium, retrieved 2026] Medium | https://medium.com/
[DevCuration, November 2026] Mosaico Raises €7.5M Seed for Robotics Data Infrastructure | https://devcuration.com/articles/mosaico-raises-e7-5m-seed-for-robotics-data-infrastructure
Articles about Mosaico
- Mosaico's Open-Source Engine Turns Robot Sensor Data Into Training Sets — The Italian startup, founded by three Ambarella veterans, raised $8.1 million to scale its on-premise platform for robotics developers.