Veeda AI

Veeda AI (legal name Veeda Innovation Inc.) is a Toronto-based artificial intelligence company building multimodal foundation world models that simulate physical reality. The goal is to let robots and other embodied agents learn by interacting with generated environments instead of the real world ^[1][2]. Computer vision researcher Sanja Fidler, previously vice president of AI research at Nvidia, founded it in 2026 with two former Nvidia colleagues, Zan Gojcic and Huan Ling ^[2][3]. Fidler is chief executive, Gojcic is chief technology officer and Ling is chief scientist ^[2][3].

Veeda was incorporated in early June 2026 and announced itself on 19 August 2026. That day The Logic reported that it had raised more than US$90 million in seed financing backed by Khosla Ventures and Radical Ventures ^[2][3][4]. Fidler describes the company's aim as building "the Matrix" for physical AI: a simulated world in which robot intelligence can interact with a virtual environment and learn from its own mistakes ^[5]. In late October 2026, the company's website described its research program but did not list a released model or product ^[1].

Overview

History

Origins at Nvidia

Fidler, a University of Toronto computer science professor, joined Nvidia in May 2018 to head a new research unit in Toronto ^[2]. BetaKit reported that she joined as director of AI at the invitation of chief executive Jensen Huang and was the Toronto lab's first employee. She later became Nvidia's vice president of AI research ^[8]. The lab was later named the Spatial Intelligence Lab, and BetaKit, citing the lab's website, reported that it grew to just over 80 members ^[2][8].

According to The Logic, the lab developed software to simulate how autonomous vehicles and humanoid agents would interact with the real world, as well as tools to generate and edit 3D content. Fidler and Ling were part of the team that created Nvidia's flagship world model for physical AI developers ^[2]. Both are listed among the authors of the January 2025 technical report on NVIDIA Cosmos, and Ling describes himself as a core contributor to Cosmos ^[9][10]. Gojcic, who joined Nvidia in November 2021 after a research internship in Fidler's Toronto lab earlier that year, led a Zurich-based team working on neural reconstruction and generative world simulation for physical AI ^[11].

On 4 August 2026, BetaKit reported that Fidler had announced on LinkedIn that she was leaving Nvidia after eight years. In the post she described her team's mission as developing "new technologies for simulation to power the creation and modeling of virtual worlds" and wrote that world models are "where the next breakthrough lies, and it is around the corner." She did not say what she would do next ^[8]. BetaKit also noted that the post listed world-model milestones at Nvidia, including a recent demonstration of real-time interactive driving in a fully AI-generated world. Fidler had given a keynote on world models for autonomous vehicles at Nvidia's GTC conference in March 2026 ^[8].

Incorporation and financing

According to corporate filings reviewed by The Logic, the three founders incorporated Veeda Innovation in early June 2026. Late in July the company issued 60.6 million seed shares at US$1 each, and on the same day it added two directors: Radical partner Tomi Poutanen and Khosla partner Sven Strohband ^[2]. The Logic noted that such changes are typically associated with fundraising ^[2].

A Swiss subsidiary, Veeda CH GmbH, was entered in the commercial register of the canton of Zurich in July 2026, with its registered seat in Uitikon. Its articles are dated 9 July 2026. The registration notice, published in the Swiss Official Gazette of Commerce on 20 July 2026, lists Veeda Innovation Inc. of Toronto as the holder of all 200 capital quotas (CHF 20,000 in total). It names Fidler as chair of management and Gojcic as managing director, each with sole signing authority ^[12]. The stated purpose of the subsidiary is developing, producing and selling hardware and software, particularly in artificial intelligence ^[12].

Launch

The Logic reported the company and its financing on 19 August 2026, citing people familiar with the matter and corporate filings ^[2]. The same day, Fidler introduced Veeda in posts on LinkedIn and X, naming Gojcic and Ling as co-founders and Khosla Ventures and Radical Ventures as investors ^[5][3]. "At Veeda, our sole mission is to build simulated reality for Physical AI. Our conviction is that this will become the critical infrastructure layer for all areas of robotics," she wrote ^[5]. Radical Ventures said the company had "just emerged from stealth" ^[7].

Founders

Fidler's academic research sits at the intersection of computer vision and graphics, with a focus on 3D vision and 3D reconstruction and synthesis ^[13]. The Logic described her as "a major figure in Canadian AI research" who holds a research chair under the federal AI strategy ^[2]. Her university homepage lists a Canada CIFAR AI Chair awarded in 2018 ^[13].

Approach

Veeda's public materials argue that "embodied intelligence will not emerge from imitation learning alone." Robots, the company says, will need to learn through interaction, much as people learn by trying, failing and trying again. It compares this with large language models, which it says gained new capabilities when training moved beyond passive learning from internet text to interactive tasks and feedback ^[1]. In the company's view, the real world is not a practical training ground for that interaction. Every mistake can carry a real physical cost. Hardware, facilities and operators do not scale the way compute does. Real-world experience also cannot be accelerated or parallelized at the scale that learning requires ^[1].

The company's answer is to have robots learn in simulated reality. Through a virtual embodiment, a robot's intelligence would interact with "an entirely virtual world in a million possible ways, learning from its own mistakes" ^[5]. Veeda says world models are the foundational technology for this, describing them as generative models that learn from large amounts of sensor and physical-world data to reach the quality, diversity and physical realism that robot training requires ^[1]. In her launch post, Fidler said the company intends to build "the foundational infrastructure for interactive robotics learning and evaluation at scale," and she asked roboticists to tell the company what they need ^[5].

Veeda's website does not describe its model architecture. In a post on Radical Ventures' website, Radical partner Tomi Poutanen wrote that Veeda processes complex sensor data alongside real-time action inputs to generate physical-world states and "automated reward signals," so that robots can learn before they are deployed ^[7]. That is an investor's description, not a technical specification published by the company.

Job postings give some detail about the planned engineering. A world-model role covers designing, training and scaling generative and predictive models across visual and spatial modalities. It lists experience with reinforcement learning, model-based learning or embodied AI as a plus ^[15]. A simulation role involves connecting Veeda's world models to simulator interfaces that customers already use, naming Isaac Lab and CARLA. It also covers building Gymnasium-style environment APIs for policy training, procedural scene and articulated asset pipelines based on OpenUSD, URDF and MJCF, and regression benchmarks that score both task success and divergence from real-hardware traces ^[16]. A robotics role covers building robot systems and data-collection workflows and running real-hardware experiments to evaluate performance ^[17]. The Logic reported that the company was also hiring staff to collect and prepare training data and to manage the chip clusters used for training ^[2].

Funding

The Logic first reported the round, putting it at more than US$90 million. Khosla confirmed its investment to the outlet, and Radical later confirmed its backing on LinkedIn ^[2]. BetaKit reported that the company confirmed it had raised US$90 million (C$124 million) from investors so far ^[3]. Radical Ventures said it co-led the investment with Khosla Ventures ^[7]. In a LinkedIn post, 2077 Ventures said it had invested as part of the "$90M+ seed round, co-led by Khosla Ventures and Radical Ventures" ^[18]. The law firm Osler, Hoskin & Harcourt said it advised Veeda on the financing ^[6].

Citing PitchBook data, The Logic said the financing would make Veeda's round one of Canada's three largest seed rounds ever ^[2]. SiliconANGLE called it one of the largest seed financings ever raised by a Canadian startup ^[4]. The reports did not disclose a valuation ^[2][3][4].

Research background

Several papers co-written by Veeda's founders at Nvidia deal with video generation, 3D reconstruction and simulation for autonomous driving. According to The Logic, Fidler and Gojcic have co-authored papers on improving AI systems' renderings of reality from different data sources, and Nvidia has patented some of Fidler's joint work with Ling on AI-generated images ^[2]. Selected publications:

Ling's website lists Difix3D+ as a CVPR 2025 oral presentation and best paper nomination ^[10].

Team and operations

Veeda calls itself "a small, deeply technical team" and says it is hiring in all four of its locations ^[1]. The company has not published a headcount. In October 2026 its job board listed member-of-technical-staff openings in world models, simulation, robotics, ML data, ML operations and ML performance, plus internships. Every posting listed Toronto as its primary location, and most also listed Zurich, Seattle or California as secondary locations ^[25]. In August 2026, The Logic reported that the company's website listed an office in Mountain View, California ^[2]. By late October the site listed the Bay Area instead ^[1].

Reception and context

Coverage of the launch placed Veeda among the research-focused startups formed by senior researchers leaving large technology companies. The Logic called Fidler "the latest star researcher to leave a U.S. tech giant to form a so-called neolab developing foundation AI models" ^[2]. BetaKit wrote that Veeda "joins a crop of neolabs building world models," and named AMI Labs and General Intuition as other examples ^[3].

Reports also pointed to the investors' earlier bets. Khosla Ventures and Radical Ventures had co-led the founding round of Discovery Loop, the AI-for-science startup led by Jeff Dean, earlier in August 2026 ^[2][4]. SiliconANGLE noted that Radical had backed World Labs and Decart ^[4]. The Logic noted that Khosla co-led a US$750 million round in January 2026 for Waabi, with Radical participating. Waabi is the autonomous-driving company led by Raquel Urtasun, another University of Toronto professor and Vector Institute co-founder, and it also develops world models ^[2]. The Logic placed Veeda among a small group of Toronto foundation model companies that includes Cohere and Ideogram ^[2].

See also

• List of neolabs
• Neolab
• World model
• Physical AI
• NVIDIA Cosmos
• Sim-to-real transfer
• Embodied AI
• Spatial intelligence
• Waabi
• World Labs

References

1. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10Veeda AI, "World models that simulate physical reality" (company website). Accessed 27 October 2026.
2. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10 ^11 ^12 ^13 ^14 ^15 ^16 ^17 ^18 ^19 ^20 ^21 ^22 ^23 ^24 ^25 ^26 ^27 ^28 ^29 ^30Murad Hemmadi, "Star researcher Sanja Fidler raises over US$90M for world model startup", The Logic, 19 August 2026 (archived copy).
3. ^1 ^2 ^3 ^4 ^5 ^6 ^7 ^8 ^9 ^10 ^11Madison McLauchlan, "Former Nvidia lab leader Sanja Fidler launches Veeda AI to tackle world models", BetaKit, 19 August 2026.
4. ^1 ^2 ^3 ^4 ^5 ^6 ^7Mike Wheatley, "Sanja Fidler's world model startup Veeda AI raises $90M in seed funding", SiliconANGLE, 19 August 2026.
5. ^1 ^2 ^3 ^4 ^5 ^6Sanja Fidler (@FidlerSanja), "I'm incredibly excited to introduce @VeedaAI, which I co-founded with my incredible longtime collaborators...", X, 19 August 2026.
6. ^1 ^2 ^3Osler, Hoskin & Harcourt LLP, "Veeda AI in its financing" (representative work), announced 19 August 2026.
7. ^1 ^2 ^3 ^4 ^5Tomi Poutanen, "World Models in Robotics", Radical Reads, Radical Ventures, 24 August 2026.
8. ^1 ^2 ^3 ^4 ^5Alex Riehl, "U of T prof Sanja Fidler leaves as Nvidia's VP of AI research", BetaKit, 4 August 2026.
9. ^1 ^2NVIDIA (Niket Agarwal et al.), "Cosmos World Foundation Model Platform for Physical AI", arXiv:2501.03575, 7 January 2025.
10. ^1 ^2 ^3 ^4Huan Ling, "Huan Ling" (personal homepage). Accessed 27 October 2026.
11. ^1 ^2Zan Gojcic, "Zan Gojcic" (personal homepage). Accessed 27 October 2026.
12. ^1 ^2Swiss Official Gazette of Commerce (SOGC), new registration notice for "Veeda CH GmbH, Uitikon (CHE-240.286.312)", Commercial Register of the Canton of Zurich, published 20 July 2026 (via Zefix).
13. ^1 ^2 ^3Sanja Fidler, "Sanja Fidler" (personal homepage, University of Toronto). Accessed 27 October 2026.
14. ^Vector Institute, "Sanja Fidler" (research team profile). Accessed 27 October 2026.
15. ^Veeda AI, "Member of Technical Staff - World Models" (job posting). Accessed 27 October 2026.
16. ^Veeda AI, "Member of Technical Staff - Simulation" (job posting). Accessed 27 October 2026.
17. ^Veeda AI, "Member of Technical Staff - Robotics" (job posting). Accessed 27 October 2026.
18. ^1 ^22077 Ventures, "We're proud to share that 2077 Ventures invested in Veeda AI as part of its $90M+ seed round...", LinkedIn, August 2026.
19. ^Jun Gao, Tianchang Shen, Zian Wang, Wenzheng Chen, Kangxue Yin, Daiqing Li, Or Litany, Zan Gojcic and Sanja Fidler, "GET3D: A Generative Model of High Quality 3D Textured Shapes Learned from Images", arXiv:2209.11163, NeurIPS 2022.
20. ^Andreas Blattmann, Robin Rombach, Huan Ling, Tim Dockhorn, Seung Wook Kim, Sanja Fidler and Karsten Kreis, "Align your Latents: High-Resolution Video Synthesis with Latent Diffusion Models", arXiv:2304.08818, CVPR 2023.
21. ^Qi Wu, Janick Martinez Esturo, Ashkan Mirzaei, Nicolas Moenne-Loccoz and Zan Gojcic, "3DGUT: Enabling Distorted Cameras and Secondary Rays in Gaussian Splatting", arXiv:2412.12507, CVPR 2025.
22. ^Jay Zhangjie Wu, Yuxuan Zhang, Haithem Turki, Xuanchi Ren, Jun Gao, Mike Zheng Shou, Sanja Fidler, Zan Gojcic and Huan Ling, "Difix3D+: Improving 3D Reconstructions with Single-Step Diffusion Models", arXiv:2503.01774, CVPR 2025.
23. ^Xuanchi Ren, Yifan Lu, Tianshi Cao et al. (including Sanja Fidler and Huan Ling), "Cosmos-Drive-Dreams: Scalable Synthetic Driving Data Generation with World Foundation Models", arXiv:2506.09042, 10 June 2025.
24. ^Sherwin Bahmani, Tianchang Shen, Jiawei Ren et al. (including Zan Gojcic, Sanja Fidler and Huan Ling), "Lyra: Generative 3D Scene Reconstruction via Video Diffusion Model Self-Distillation", arXiv:2509.19296, 23 September 2025.
25. ^Veeda AI, "Veeda AI Jobs" (job board). Accessed 27 October 2026.