Can Dyna's foundation model generalize across customer sites without extensive on-site engineering, or does each deployment still require substantial custom work?

In its own 2026 blog post reflecting on its first year of production deployments, Dyna acknowledged that getting from a signed customer to a robot 'running reliably without us' took weeks to months of on-site engineering per site, and that this work 'didn't transfer cleanly to the next customer' — a gap between its stated out-of-the-box vision and what it has observed in practice so far.

How many paying, named production customers does Dyna actually have, and at what scale?

Dyna has publicized deployments in hotels, restaurants, laundromats, and gyms and cited a single longest-running deployment of about 10 months as of its most recent public update, but has not named specific commercial customers or disclosed revenue; some third-party trackers explicitly flag its deployment posture as pilot-stage rather than at-scale.

Dyna Robotics

Dyna Robotics builds general-purpose dual-arm robots powered by a proprietary embodied AI foundation model, deployed in hotels, restaurants, laundromats, and gyms.

Last updated: July 15, 2026

Summary

Dyna Robotics builds general-purpose, dual-arm robots powered by a proprietary embodied AI foundation model called DYNA-1, aimed at automating repetitive manual tasks — such as folding, food preparation, and packaging — in businesses like hotels, restaurants, laundromats, and gyms. Founded in 2024 by repeat entrepreneurs Lindon Gao and York Yang, who previously sold smart-cart startup Caper AI to Instacart for $350 million, and former Google DeepMind researcher Jason Ma, the company raised roughly $143.5 million within about a year of launching, at a valuation exceeding $600 million.

Company Story

Dyna Robotics was founded in 2024 by Lindon Gao and York Yang — co-founders of Caper AI, an AI-powered smart shopping cart startup Instacart acquired for $350 million in 2021 — together with Jason Ma, a former Google DeepMind and Meta AI research scientist who completed his Ph.D. at the University of Pennsylvania’s GRASP Laboratory. According to Dyna, Gao and Yang teamed up with Ma after seeing firsthand, while scaling Caper’s hardware-and-software product, how quickly AI systems could improve once deployed in the real world, and wanted to apply the same approach to general-purpose robotics (The Robot Report).

The company emerged from stealth in March 2025 with a $23.5 million seed round co-led by CRV and First Round Capital, at a reported valuation of roughly $100 million. The following month, Dyna launched DYNA-1, a robotics foundation model the company said could push its dual-arm robots to a success rate above 99% across 24 hours of continuous, non-stop operation. Within six months, Dyna said its robots were running up to 16 hours a day in commercial settings including hotels, restaurants, laundromats, and gyms, performing tasks such as folding napkins and towels. In September 2025, the company closed a $120 million Series A led by Robostrategy, CRV, and First Round Capital, with participation from Salesforce Ventures, NVIDIA’s NVentures, the Amazon Industrial Innovation Fund, Samsung Next, and LG Technology Ventures, at a post-money valuation exceeding $600 million (PR Newswire; Sacra).

In a 2026 blog post reflecting on its first year of production deployments, Dyna offered an unusually candid account of the gap between its “out-of-the-box” pitch and the reality of scaling a foundation-model robot across customer sites: its longest-running deployment had reached about 10 months of daily use, but getting each new site from a signed contract to a robot “running reliably without us” required weeks to months of on-site engineering that “didn’t transfer cleanly to the next customer.” The company said this experience led it to bring hardware design and manufacturing in-house alongside its model research, arguing that vertical integration — while more capital-intensive — was the only way to close the loop between deployment data and model improvement (DYNA Robotics blog).

Founders

Lindon Gao co-founded Dyna Robotics and serves as CEO. A graduate of Stuyvesant High School in New York, he studied Finance and International Business at NYU Stern School of Business and worked as an investment banking analyst at Goldman Sachs before turning to entrepreneurship. He co-founded QueueHop, an anti-theft apparel-tag startup backed by Y Combinator, before co-founding Caper AI in 2016 with York Yang and others; Caper’s AI-powered smart shopping carts were deployed at retailers including Kroger, ShopRite, and Aldi before Instacart acquired the company for $350 million in 2021. Gao then served as VP and General Manager of In-Store at Instacart before leaving to co-found Dyna in 2024 (Yespress; Wikipedia).

York Yang co-founded Dyna Robotics and serves as its CTO. He holds a bachelor’s degree in Electronic Engineering from Zhejiang University and a master’s degree in Computer Science from UCLA, and served as CTO of Caper AI before continuing as a Principal Research Engineer at Instacart following its acquisition of Caper. He is Gao’s longtime collaborator, having also worked with him on QueueHop before Caper (The Robot Report).

Jason Ma (Yecheng Ma) co-founded Dyna Robotics and serves as Chief Scientist. He earned an undergraduate degree in Computer Science from Harvard and a Ph.D. in Computer and Information Science from the University of Pennsylvania’s GRASP Laboratory, advised by Dinesh Jayaraman and Osbert Bastani, where his research focused on reinforcement learning and foundation models for robot learning. He was named an Apple Scholar in AIML during his doctorate and was a lead author of NVIDIA’s Eureka reward-design research. Before Dyna, he worked as a researcher at Google DeepMind, NVIDIA AI, and Meta AI, and his work has been recognized with an RSS Pioneers award, an OpenAI Fellowship, a Best Paper Finalist nod at ICRA 2024, and a spot among NVIDIA Research’s top 10 projects of 2023 (GRASP Lab, University of Pennsylvania; Jason Ma personal website).

Products

Dyna’s core product is DYNA-1, a full-stack embodied AI system built around two industrial robotic arms mounted on a compact wheeled base designed to integrate with existing workstations. The arms use quick-swap grippers — including suction, parallel, and custom end-effectors — that let the same hardware handle a range of materials, from napkins and towels to food containers, without task-specific reprogramming. Dyna says its foundation model lets robots learn new skills within hours rather than weeks and improve fleet-wide as each deployment generates new training data. The company markets DYNA-1 across factory automation (kitting, assembly, pick-and-place, and packaging), laundry and garment handling (folding at commercial scale), and food and hospitality automation (napkin folding, plating, and related food-service tasks); it has since previewed a research iteration called DYNA-1i, aimed at more open-world dexterity (Dyna Robotics; Sacra).

Funding

Dyna Robotics has raised approximately $143.5 million across two rounds since its 2024 founding. It closed a $23.5 million seed round in March 2025, co-led by CRV and First Round Capital, followed just six months later by a $120 million Series A in September 2025 led by Robostrategy, CRV, and First Round Capital, with participation from Salesforce Ventures, NVIDIA’s corporate venture arm NVentures, the Amazon Industrial Innovation Fund, Samsung Next, and LG Technology Ventures. The Series A valued the company at more than $600 million post-money (PR Newswire; Sacra).

Partnerships

Dyna has not publicly named specific commercial customers, describing its deployments in aggregate as spanning hotels, restaurants, laundromats, and gyms. Its investor base includes several large technology and industrial corporations with strategic interests in robotics and automation — NVIDIA (via NVentures), Amazon (via its Industrial Innovation Fund), Samsung, LG, and Salesforce — any of which could represent potential future commercial or technical collaboration, though no formal partnerships with these investors have been announced beyond their funding participation.

Timeline

2024: Lindon Gao, York Yang, and Jason Ma found Dyna Robotics.
March 2025: Dyna emerges from stealth with a $23.5 million seed round.
April 2025: Dyna launches its DYNA-1 robotics foundation model.
September 2025: Dyna closes a $120 million Series A at a valuation exceeding $600 million.
2025: Dyna previews DYNA-1i, an “Open-World Dexterity” research iteration of its model.
2026: Dyna publishes a candid account of its first year of production deployments, citing a roughly 10-month longest-running customer engagement and ongoing per-site engineering challenges.

Market Context

Dyna competes in the fast-growing field of robotic foundation models and general-purpose manipulation, alongside companies such as Physical Intelligence, Skild AI, and Covariant, as well as more vertically focused kitchen-automation players like Miso Robotics. Its pitch centers on a single foundation model that can generalize across many manipulation tasks and environments with minimal customization, a claim that distinguishes it from narrower single-task automation vendors but that the company itself has acknowledged remains only partially proven in production. Dyna’s founders’ prior exit — the $350 million sale of Caper AI to Instacart — and the rapid, corporate-investor-heavy fundraising that followed Dyna’s 2025 launch reflect strong venture interest in embodied AI foundation models generally, even as the company has been candid about the operational difficulty of scaling deployments beyond initial pilot sites.

Source Notes

This profile draws on Dyna Robotics’ own website and blog (including a notably candid 2026 post on the operational challenges of its first year in production), funding announcements from PR Newswire and The Robot Report, funding and valuation data from Sacra, Wikipedia’s entry on the founders’ prior company Caper AI, a Yespress profile of CEO Lindon Gao, and University of Pennsylvania GRASP Lab and personal-website biographies of co-founder Jason Ma. Because Dyna is an early-stage private company that has not named specific paying customers or disclosed revenue, claims about deployment scale and commercial traction in this profile should be treated as company-reported rather than independently verified, and re-checked against more recent company statements for anything time-sensitive.

Funding Table

Date	Round	Amount	Valuation	Investors	Source
2025-03	Seed	$23.5 million	Unknown	CRV, First Round Capital	Dyna Robotics closes $120M funding round to scale robotics foundation model — The Robot Report
2025-09	Series A	$120 million	over $600 million post-money	Robostrategy, CRV, First Round Capital, Salesforce Ventures, NVentures (NVIDIA), Amazon Industrial Innovation Fund, Samsung Next, LG Technology Ventures	Dyna Robotics Raises $120 Million to Advance Robotic Foundation Models — PR Newswire

Structured Timeline

2024Founded by repeat Caper AI founders and a DeepMind alumnus
Lindon Gao and York Yang, who previously sold smart-cart startup Caper AI to Instacart for $350 million, founded Dyna Robotics with Jason Ma, a former Google DeepMind research scientist and GRASP Lab Ph.D.
2025-03Emerges from stealth with $23.5 million seed round
Dyna announced a $23.5 million seed round co-led by CRV and First Round Capital, at a reported valuation of roughly $100 million.
2025-04Launches DYNA-1 foundation model
Dyna released DYNA-1, a robotics foundation model the company said could push dual-arm robots to a 99%-plus success rate across 24 hours of continuous operation.
2025-09Closes $120 million Series A
Dyna raised a $120 million Series A led by Robostrategy, CRV, and First Round Capital, with participation from Salesforce Ventures, NVIDIA's NVentures, and other corporate investors, at a post-money valuation exceeding $600 million.
2025Announces DYNA-1i (Open-World Dexterity)
Dyna previewed DYNA-1i, a research iteration of its foundation model aimed at handling greater environmental variability and more complex manipulation skills beyond its initial deployments.

Founder Profiles

Open Questions

These are research questions readers, analysts, and AI systems may want to verify as the company develops.

Can Dyna's foundation model generalize across customer sites without extensive on-site engineering, or does each deployment still require substantial custom work?
In its own 2026 blog post reflecting on its first year of production deployments, Dyna acknowledged that getting from a signed customer to a robot 'running reliably without us' took weeks to months of on-site engineering per site, and that this work 'didn't transfer cleanly to the next customer' — a gap between its stated out-of-the-box vision and what it has observed in practice so far.
How many paying, named production customers does Dyna actually have, and at what scale?
Dyna has publicized deployments in hotels, restaurants, laundromats, and gyms and cited a single longest-running deployment of about 10 months as of its most recent public update, but has not named specific commercial customers or disclosed revenue; some third-party trackers explicitly flag its deployment posture as pilot-stage rather than at-scale.