Robotec.ai

Robotec.ai Deep Dive: From the RoSi Multi-Engine Physics Simulation Platform to the SLINGSHOT Real-to-Sim Pipeline and the Open-Source RAI Embodied-Agent Framework

Robotec.ai (Warsaw, 2019) builds Physical-AI simulation and digital twins: the RoSi multi-engine simulation platform, the SLINGSHOT real-to-sim pipeline, GPU Sensors, and the open-source RAI agent framework, co-authoring the ROS 2 Simulation Interfaces standard with NVIDIA and Intrinsic.

Executive Summary

Robotec.ai (legal entity Robotec.ai sp. z o.o.) is a deep-tech company founded in 2019 and headquartered in Warsaw, Poland. Its stated mission is to accelerate the development, validation and deployment of autonomous systems and Physical AI through high-fidelity, open simulation platforms. Rather than building a single physics solver, Robotec.ai positions itself as the simulation and digital-twin layer for Physical AI: it orchestrates third-party engines (NVIDIA Isaac Sim, Open 3D Engine / O3DE, Genesis) under one product surface, turns real-site scans into OpenUSD digital twins through its SLINGSHOT pipeline, and adds an open-source agentic AI framework (RAI) on top. The company serves warehouse logistics, manufacturing, mining, agriculture and autonomous-driving customers, and it co-authored the ROS 2 Simulation Interfaces standard with NVIDIA and Intrinsic, now implemented in Isaac Sim 5.0, O3DE 2505 and Gazebo. This report assembles the public record on team, technology, capital, commercial evidence and partnerships, and rules item by item on the six groups of conflicting public accounts recorded in DISCREPANCIES. Data is current as of 14 November 2026.

Robotec.ai is privately held with no disclosed priced venture rounds. Third-party databases describe it as accelerator-backed (one names AI Venture Labs) while another states it has never raised funding; the company itself publishes no financing figures. Headcount is reported between roughly 50 and 72 across sources, with one weekly dataset showing a peak near 72 in late 2024 and about 53 by September 2026, a 17% year-over-year decline. These are precisely the points where public accounts disagree, and this report flags them rather than averaging them away. Technically the company is credible: it leads open robotics standards, ships multiple products, and demonstrates real hardware-in-the-loop integration with AMD, Qualcomm, AWS, Analog Devices and Tier IV / Autoware. Commercially it is a recognised but small player whose pricing is quote-only and whose revenue is undisclosed.

Company Overview

Robotec.ai was founded in 2019 in Warsaw and is consistently described as a deep-tech company focused on robotics simulation, digital twins and Physical AI. The legal entity is Robotec.ai sp. z o.o., a Polish limited liability company. Leadership is described across sources as Michał Niezgoda (CEO and co-founder), Adam Dąbrowski (Chief Technology Officer), Marcin Majewski (Chief Financial Officer and co-founder) and Willem Bouwmeester (Chief Commercial / Science Officer and co-founder), with Ruben Kragten as Vice President of Global Sales and Business Development and Wojciech Jaworski as Director of Product. The co-founder set varies by source, which is one of the discrepancies ruled in DISCREPANCIES. The company presents itself as "the Simulation and Digital Twin layer for Physical AI" and stresses human-centred robotics, open-source standards and ROS 2 middleware as its technical backbone.

Headcount is the weakest-documented quantitative claim. Crustdata's proprietary weekly dataset tracked Robotec.ai from about 41 employees in October 2022 to roughly 72 by October 2024, then to about 53 by September 2026, a 17.2% year-over-year decline; its Growth Score was 38/100 (steady growth) at that point, with web traffic up 56% quarter-over-quarter but only one open role. Village.ai lists "50+", Prospeo lists 51–100, RocketReach lists 49–53, and PitchBook lists 59. The earlier company-list entry on this site used "approximately 70" based on the 2024 peak; this report revises that to "approximately 50–60, with a 2024 peak near 72" and treats the exact number as unverified. Crunchbase presents funding-event entries that are clearly placeholder ("Lorem ipsum"), so it is excluded from the financing account.

Product and Technology

Robotec.ai's product surface is organised around one orchestration platform and several specialised modules. RoSi is the flagship digital-twin simulation platform for Physical AI: a full physics-based simulation environment for validating autonomous systems and robotics stacks, spanning single sensors to full robot fleets, in software-in-the-loop (SiL) and hardware-in-the-loop (HiL) setups. RoSi integrates simulation engines (Isaac Sim, O3DE, Genesis), a web GUI scenario builder, a backend and REST API for headless launches and CI/CD hooks, and metrics and analytics that record rosbags and let users replay exact runs by seed and asset version. AIKROS is the operational digital twin for warehouse and logistics, a 2D process simulator that runs up to roughly 10,000× speed with "what-if" analysis of throughput, bottlenecks and strategy. SLINGSHOT is the automated real-to-sim pipeline that turns site scans, video and sensor logs into a SimReady 3D twin delivered in OpenUSD, advertised as a seven-day turnaround from data delivery to a working twin, with a "Living Digital Twin" subscription that keeps the twin in sync with the real site.

GPU Sensors is Robotec.ai's open-source, GPU-accelerated lidar and sensor simulation library, delivering realistic point clouds at scale for perception testing and synthetic-data generation; RoSi Sensors, powered by the RGS ray-tracing library, is the productised form used with AMD. RAI is the company's open-source, model-agnostic agentic framework for embodied AI on ROS 2 (covered in depth in the companion tool entry at /en/tools/rai/), and is one of the flagship projects of the ROS 2 Embodied AI Community Group. The technology page also mentions AWSIM, a Unity-based open simulator for Autoware used to develop and validate autonomous-driving scenarios. The architectural bet is explicit: Robotec.ai rarely invents a new physics core; instead it wraps, orchestrates and standardises the best available engines behind a ROS 2-native, OpenUSD-centred interface, so a customer's twin loads into RoSi, Isaac Sim or O3DE without conversion or lock-in.

Simulation Stack and Ecosystem Partnerships

The simulation stack is deliberately multi-engine. Within RoSi, Isaac Sim runs alongside O3DE to enable cross-engine interoperability and reuse of Isaac-based assets and pipelines; Genesis is used for massively parallel reinforcement learning and humanoid simulation, and with AMD the company trained humanoid locomotion policies on AMD Instinct GPUs and deployed them on Ryzen AI edge hardware. OpenUSD is the scene format across the stack: SLINGSHOT emits every twin as an OpenUSD scene. AWS is one of the execution backends for RoSi, enabling scalable parallel simulation without tying the platform to a single infrastructure. On the robotics side, over half of Robotec.ai's customers run ROS 2; the company builds Nav2-based navigation pipelines and MoveIt manipulation workflows and connects both to RAI for intent-driven, higher-level planning. It also supports proprietary middleware.

The partnership story is the company's strongest strategic asset. Robotec.ai co-authored the ROS 2 Simulation Interfaces standard with NVIDIA and Intrinsic, an open interface (common message, service and action types) that is implemented in NVIDIA Isaac Sim 5.0, O3DE 2505 and Gazebo (adoption planned for October 2025), lowering the cost of moving projects between simulators. With NVIDIA and Intrinsic it led a case study standardising simulation interfaces; with AMD and AMD Silo AI it optimised its simulation stack for AMD Instinct GPUs and the ROCm software stack and tested the first fully autonomous warehouse robot on AMD Ryzen AI; with AMD and Liquid AI it built an edge agentic robotics demo running Liquid's LFM2-VL vision-language model, orchestrated by RAI; with Qualcomm it built a photorealistic retail simulation for a mobile manipulator with hardware-in-the-loop validation on Qualcomm edge hardware; with AWS it demonstrated cloud-native, scalable robotics simulation on O3DE; with Analog Devices (ADI) it delivered a digital twin of a semiconductor production bay; and with Tier IV / Autoware it improved autonomous-driving simulation realism and sensor validation. These are vendor demonstrations, not independent blind benchmarks.

Business Model, Customers and Go-to-Market

Robotec.ai sells through three routes. "Products / Quick start" licenses the full simulation platform or individual modules for leading sim engines (O3DE, Isaac Sim, Genesis) so the customer's own team drives. "Engineering projects" are fixed-scope statements of work: build the twin, custom validation, sim-to-real, train Physical AI. "Dedicated simulation teams" are co-development engagements where Robotec.ai engineers take ownership of outcomes under the customer's requirements and SOWs. Pricing is quote-only; the company does not publish a price list, a free tier or a self-serve trial, which the AI Tech Suite profile lists among its cons. The target buyers are robotics engineers, autonomous-vehicle developers and industrial operators in warehouse logistics, mining, agriculture, semiconductor manufacturing and automotive.

Named collaborations function as both evidence and channel. Qualcomm, AWS, AMD, NVIDIA, Intrinsic, Analog Devices, Tier IV / Autoware, Booz Allen Hamilton (Bill Vass, CTO, publicly endorsed Robotec.ai) and Open Robotics (Geoffrey Biggs, CTO, also endorsed it) all appear in Robotec.ai material. These are partner demonstrations and endorsements rather than disclosed commercial contracts with values, so they evidence technical credibility and ecosystem access more than booked revenue. The company's own case-study pages describe outcomes (closing the sim-to-real gap, confirming real-time performance on edge hardware, validating automation scenarios) but do not publish contract sizes, recurring revenue or customer counts. For a private company without audited financials, the honest reading is that Robotec.ai has shipped products, won marquee partners and built a credible engineering brand, while its revenue base, unit economics and pipeline value remain undisclosed.

Competitive Landscape

Robotec.ai sits in the robotics-simulation and digital-twin layer, a crowded but fragmented space. Against game-engine simulators it competes with NVIDIA Isaac Sim and the broader Omniverse stack, with O3DE (where it is itself a lead contributor), and with Gazebo; against cloud and multi-physics CaaS it is adjacent to the simulation orchestration offered by firms such as Rescale and SimScale, though those target engineering HPC rather than robot fleets. Against embodied-AI frameworks it is adjacent to Physical Intelligence, Skild AI and World Labs on the model side, and to NVIDIA's Cosmos and Isaac on the platform side. Robotec.ai's differentiator is not a novel solver but standards leadership and integration breadth: co-authoring the ROS 2 Simulation Interfaces standard and wrapping Isaac Sim, O3DE and Genesis behind one ROS 2-native, OpenUSD-centred surface.

The comparison that matters for this list is with NVIDIA. NVIDIA supplies the engines (Isaac Sim, Cosmos, Omniverse) and the silicon; Robotec.ai supplies the orchestration, the real-to-sim twin generation and the agentic layer (RAI) on top, and it co-wrote the interoperability standard with NVIDIA rather than against it. This is a complementarity relationship more than a head-to-head one. Against pure-play digital-twin vendors, Robotec.ai's strength is its Physical-AI orientation (synthetic data, SiL/HiL, agentic planning) rather than static visual twins. Its principal constraints are scale and capital: it is a roughly 50–60 person Warsaw team with no disclosed large financing, competing for mind-share and deals against far better-funded platforms while depending on partners (NVIDIA, AMD, AWS, Qualcomm) for the underlying engines and reach.

Financing and Capital Structure

Robotec.ai is privately held and has disclosed no priced venture rounds. PitchBook lists it as "Accelerator/Incubator Backed" with a single investor, AI Venture Labs, and a status of "Generating Revenue", while Prospeo states the company "has never raised funding before." Crunchbase shows funding-event rows that are placeholder text ("Lorem ipsum"), so they are discarded. The company itself publishes no financing history, valuation or revenue. The defensible reading is that Robotec.ai is effectively bootstrapped-to-revenue with, at most, one accelerator relationship, and that no public round amount, valuation or post-money figure should be cited. This is recorded as a discrepancy in DISCREPANCIES because the "accelerator-backed" and "never raised" accounts cannot both be fully true, and because the absence of any company disclosure leaves the question open.

The capital-structure consequence is straightforward. Without external venture capital, Robotec.ai's growth is funded by services revenue and partner co-development rather than a financed runway; this keeps it independent and avoids single-investor concentration, but it also caps the scale of any go-to-market or R&D push relative to better-capitalised competitors. For diligence, the items that would carry weight are a company statement on financing and investors, signed partner contracts with values, and disclosed revenue or headcount trajectory; none of these is currently public. The earlier company-list entry described Robotec.ai as private with undisclosed financing and no public rounds, which remains the accurate summary.

DISCREPANCIES

1. Headcount and its trajectory. Account A: Crustdata's weekly proprietary dataset tracked Robotec.ai from about 41 employees in October 2022 to roughly 72 by October 2024, then to about 53 by September 2026, a 17.2% year-over-year decline, with one open role and a Growth Score of 38/100. Account B: Village.ai says "50+", Prospeo says 51–100, RocketReach says 49–53, PitchBook says 59. Account C: this site's earlier company-list entry used "approximately 70" based on the 2024 peak. Ruling: report headcount as approximately 50–60 with a 2024 peak near 72, treat the exact current number as unverified, and note the decline shown by the one source that tracks it weekly; do not present any single figure as authoritative. Source grade: PROPRIETARY DATASET and DATABASE versus this site's prior peak-based estimate.

2. Financing status. Account A: PitchBook lists Robotec.ai as "Accelerator/Incubator Backed" with one investor, AI Venture Labs. Account B: Prospeo states the company "has never raised funding before." Account C: Crunchbase shows funding-event rows that are placeholder "Lorem ipsum" text and are discarded. Ruling: describe Robotec.ai as privately held and effectively bootstrapped-to-revenue, with at most one accelerator relationship, and explicitly state that no priced round, valuation or revenue figure is publicly disclosed; flag the accelerator-backed versus never-raised conflict as unresolved. Source grade: DATABASE versus DATABASE and discarded placeholder.

3. Co-founder and leadership composition. Account A: Robotec.ai material and RocketReach list Michał Niezgoda (CEO and co-founder), Adam Dąbrowski (CTO), Marcin Majewski (CFO and co-founder) and Willem Bouwmeester (CSO/CCO and co-founder). Account B: some profiles list Bouwmeester as CCO rather than emphasising co-founder status, and Village.ai's people list omits him from the founder label while listing others. Ruling: report the verified leadership as Niezgoda (CEO, co-founder), Dąbrowski (CTO), Majewski (CFO, co-founder) and Bouwmeester (co-founder, commercial/science), and mark the precise co-founder set as varying by source rather than settling it. Source grade: COMPANY and PROFILES versus PROFILES.

4. Legal entity and brand naming. Account A: Crunchbase gives the legal name as Robotec.ai sp. z o.o., consistent with a Polish limited-liability company. Account B: code repositories and community pages use "RobotecAI" (no dot) as the GitHub organisation and in paper authorship, while the commercial brand is "Robotec.ai". Ruling: use Robotec.ai sp. z o.o. as the legal entity and Robotec.ai as the brand, and note that "RobotecAI" is the open-source / repository spelling; these are the same company, not two. Source grade: DATABASE and OPEN-SOURCE versus BRAND.

5. RAI's ownership and positioning. Account A: Robotec.ai describes RAI as "our open-source agentic AI framework for ROS 2" and the technology page presents it as the company's own framework. Account B: the RAI repository and community pages describe RAI as "one of the main projects in focus of the Embodied AI Community Group," a ROS 2 community effort, while also being maintained by Robotec.ai. Ruling: record RAI as an open-source framework created and maintained by Robotec.ai that is also a flagship project of the ROS 2 Embodied AI Community Group; the dual attribution is accurate and not a contradiction, but the "company-owned" versus "community-owned" framing should not be overstated. Source grade: COMPANY and OPEN-SOURCE.

6. Vendor claims of scale and autonomy. Account A: Robotec.ai states "over half of our customers run ROS 2" and, with AMD and Liquid AI, demonstrated "the first fully autonomous warehouse robot powered exclusively by AMD Ryzen AI processors." Account B: these are the company's own claims and partner demonstrations, not independently audited benchmarks or disclosed customer contracts. Ruling: report both claims as company-side statements, attribute them to Robotec.ai and its partners, and mark them as unverified by any independent source; do not present the autonomous-warehouse claim as a certified industrial deployment. Source grade: COMPANY and PARTNER versus independent audit (absent).

Risks and Due Diligence Gaps

The first gap is the absence of disclosed financials. There are no audited statements, no revenue, no contract values, no customer counts and no financing history. For a company whose partners include NVIDIA, AMD, AWS and Qualcomm, the gap between ecosystem access and disclosed commercial traction is wide, and any revenue estimate would be speculation. Diligence should request signed contracts with values, recurring-revenue evidence and a headcount plan before treating Robotec.ai as a scaled vendor.

The second gap is the orchestration-versus-solver dependency. Robotec.ai's value is integration and standards, not a proprietary physics core; its platform depends on Isaac Sim, O3DE and Genesis remaining available, licence-stable and performant. A change in any engine's licensing, road-map or performance could materially affect RoSi. The OpenUSD and ROS 2 standard leadership mitigates lock-in for customers, but it does not remove Robotec.ai's own dependence on upstream engines.

The third gap is the headcount trajectory. The one source that tracks headcount weekly shows a decline of about 17% year-over-year into late 2026, even as web traffic grew. A shrinking team at a services-and-engineering-heavy company can signal project concentration, margin pressure or a pivot; it is the opposite of the growth signal a financing round would imply. This should be reconciled with the company before any long-term commitment.

The fourth gap is validation maturity. Robotec.ai's evidence is overwhelmingly partner demonstrations and case studies (Qualcomm, AWS, AMD/Liquid AI, ADI, Tier IV), not independent blind benchmarks with published numbers. RAI ships a benchmarking suite (rai_bench), but the company does not publish a single headline benchmark. Physical-AI buyers need repeatable, comparable metrics; vendor demos are a starting point, not proof.

The fifth gap is partnership concentration versus independence. Robotec.ai's credibility rests heavily on NVIDIA, AMD, AWS and Qualcomm relationships. Those relationships are strategic assets, but they also mean Robotec.ai's visibility rises and falls with its partners' road maps and priorities. A private, lightly funded company in that position should be assessed for both opportunity (standards leadership) and fragility (dependence on larger platforms).

The sixth gap is classification for this list. Robotec.ai is a simulation-and-digital-twin company whose Physical-AI relevance comes from synthetic data, SiL/HiL validation and the RAI agentic layer. It is not a CAE surrogate-model vendor and not a world-model lab; its objective function is robot-validation throughput and sim-to-real closure, not verifiable numerical prediction of physical fields. This report records that boundary explicitly so Robotec.ai is not mistaken for a neural-solver company.

Overall Assessment

Technical strength scores 7.5. Robotec.ai leads open robotics standards (the ROS 2 Simulation Interfaces standard), ships a multi-engine orchestration platform (RoSi), a real-to-sim pipeline (SLINGSHOT), GPU-accelerated sensor simulation and an open-source agentic framework (RAI), and demonstrates real hardware-in-the-loop work with AMD, Qualcomm and AWS. The deduction reflects that its core is integration rather than a proprietary physics solver, so deep-physics novelty is limited.

Commercial capability scores 7.0. The company has marquee partners, shipped products, three clear go-to-market routes and public endorsements from senior robotics figures. The deduction reflects quote-only pricing with no published price list or trial, undisclosed revenue, and evidence that is partner demonstrations rather than booked, valued contracts.

Financing scores 4.5. Robotec.ai is privately held with no disclosed priced rounds and at most one accelerator relationship; it is effectively bootstrapped-to-revenue. The deduction reflects the absence of any public capital, valuation or financing history, which caps scalable go-to-market relative to funded competitors.

Product maturity scores 8.0. RoSi, AIKROS, SLINGSHOT, GPU Sensors and RAI are all released or open source, with documented modules, APIs and case studies across warehouse, manufacturing, mining, agriculture and autonomous driving. The deduction reflects that several offerings are demonstration- and services-led, with pricing and SLAs disclosed only under NDA.

Market position scores 7.0. Robotec.ai is a recognised open-standards leader in robotics simulation, with NVIDIA/AMD/AWS/Qualcomm ecosystem access that far exceeds its headcount. The deduction reflects intense competition from far better-capitalised platforms and a dependence on partners for engines, silicon and reach.

The composite is 6.8. Robotec.ai is a credible, standards-leading Physical-AI simulation and digital-twin company whose principal assets are ecosystem partnerships, open-standards authorship and a multi-engine orchestration surface, and whose principal liabilities are undisclosed finances, a declining headcount signal, solver dependence and demo-grade validation. The signals worth tracking, in order, are a first independent benchmark or named industrial deployment with results, a company financing or revenue disclosure, reconciliation of the headcount trajectory, and the production adoption of the ROS 2 Simulation Interfaces standard beyond NVIDIA, O3DE and Gazebo.

References

  1. Robotec.ai official site, robotec.ai: company positioning as "the Simulation and Digital Twin layer for Physical AI," product loop (RoSi, AIKROS, SLINGSHOT), ways to work (products, engineering projects, dedicated teams).
  2. Robotec.ai technology page, robotec.ai/technology: RoSi orchestration, GPU Sensors, SLINGSHOT pipeline, Isaac Sim / O3DE / Genesis, AWSIM, OpenUSD, ROS 2 standard co-authored with NVIDIA and Intrinsic, Nav2, MoveIt, RAI, LangChain, AMD/Liquid AI edge demo.
  3. RoSi documentation, rosi-docs.robotec.ai/docs/intro: digital-twin integration, synthetic-data collection, productivity evaluation, software-in-the-loop validation, simulator engine, web GUI, backend and API, metrics and analytics, scenario scale.
  4. SLINGSHOT page, robotec.ai/slingshot: real-to-sim pipeline, seven-day twin delivery, OpenUSD output, Living Digital Twin subscription, standalone and connected integration.
  5. RAI repository, github.com/RobotecAI/rai, README, components (rai core, rai whoami, rai_asr, rai_tts, rai_sim, rai_bench, rai_perception, rai_finetune, rai_nomad) and citation.
  6. Rachwał K., Majek M., Boczek B., Dąbrowski K., Liberadzki P., Dąbrowski A., Ganzha M., "RAI: Flexible Agent Framework for Embodied AI," arXiv:2505.07532, 2025 (submitted 12 May 2025; PAAMS'25).
  7. Robotec.ai case study, "A new standard for open-source robotics simulation" (NVIDIA x Intrinsic): ROS 2 Simulation Interfaces standard, adoption in Isaac Sim, O3DE, Gazebo.
  8. Robotec.ai blog, "standardizing simulation interfaces for open-source robotics" and "embraces open standards": standard details, Isaac Sim 5.0 / O3DE 2505 / Gazebo adoption, OpenUSD adoption, ROSCon 2025 talk.
  9. AMD blog, "Advancing Robotics Simulations with Robotec.ai and AMD Silo AI" (2025): RoSi on AMD Instinct / ROCm, RoSi Sensors with RGS ray tracing, open-source ecosystem, autonomous warehouse robot on Ryzen AI.
  10. Robotec.ai blog, "advancing agentic AI with Liquid AI and AMD" and Qualcomm collaboration post: LFM2-VL on-device VLM, RB-KAIROS+ HiL on Ryzen, CES 2026 humanoid retail demo, sim-to-real on Qualcomm edge hardware.
This report is compiled from public sources and does not constitute investment or procurement advice. Robotec.ai is a private, unlisted company with no audited financials and no disclosed revenue, financing or valuation; headcount, financing status, co-founder composition, legal-name spelling, RAI ownership and vendor scale claims carry six groups of conflicting public accounts (see DISCREPANCIES). Uncertainties are flagged throughout; please rely on the company's primary announcements, signed partner contracts, peer-reviewed papers and independent benchmarks.

This report is compiled from public sources. It is for reference only and does not constitute investment advice. Data is current as of the publication date; please refer to the company’s official filings.