Rhoda AI Uses Internet-Scale Video to Train Its Direct Video Action Model for Robots
- Published
- Source
- RoboticsBusinessReview.com
Rhoda AI is using internet-scale video to pretrain its direct video action (DVA) model for robotics. The company emerged from 18 months in stealth earlier this year with $450 million in Series A funding and is working with customers in automotive, manufacturing, logistics, and e-commerce. Co-founder and CEO Jagdeep Singh said the approach aims to address the shortage of large-scale physical-world training data, which he described as a constraint on robust robot deployment.
Rhoda AI says its DVA model learns from video frames, action tokens, and language tokens, predicts what may happen over the next few hundred milliseconds, and converts that prediction into robot actions. The robot executes the actions, then the model uses a new observation to make another prediction, closing the loop. The company recently extended the model with mobile DVA, which it said combines manipulation and mobility in one policy. Singh argued that broad video exposure can help models handle varied real-world conditions better than training on small, narrowly collected datasets.
Rhoda AI’s disclosed monitoring point is whether its video-pretrained DVA and mobile DVA models can perform robustly in real-world customer deployments across the named sectors. The source describes the company as working with customers to deploy DVA, but does not provide deployment results, customer names, or a timeline for broader availability. Singh also said the number of autonomous manipulation-capable robots solving real-world problems was effectively zero; no independent validation of that assessment was provided.
Original report
How Rhoda AI is using video data to create generalized AI models for robots
RoboticsBusinessReview.com · Brianna WesslingThis briefing is an independently written summary based on publicly available reporting and is provided for industry information and news discovery. The original report and source publication are credited and linked where applicable. RobotToday does not claim ownership of third-party source material.
Rights concerns? Contact [email protected] with the relevant URL and details. We will review the matter and take appropriate action where warranted.
More related news
- 1
- 2
- 3
- 4
-
5
RobotToday Weekly November 2 – 6, 2026 RobotToday · 6 days ago
- 6
- 7
- 8
- 9
- 10
- 11
- 12
Related Suppliers
Rhoda AI
United States
Palo Alto robotics company building a video-prediction robot control model (DVA, FutureVision) and a wheeled 25 kg-payload robot; USD 450M Series A in 2026.
EA Mobile Robotics
Sweden
Mobile-robot unit of Swedish integrator Elektroautomatik; TC-series AGVs from 1,000 to 6,000 kg, built in Skovde and running since 1986
MFR (Mobile Fulfillment Robotics) — Samsung C&T Venture
Korea, Republic of (South)
Samsung C&T robotics initiative for mobile fulfillment and autonomous delivery robots deployed in residential complexes and logistics sites in South Korea.
Shenzhen Liangyuan Innovation Technology Co., Ltd. 深圳亮源新创科技有限公司
China
Shenzhen embodied-AI lab led by ex-OpenAI researcher Roger Jiang, building Physical AI foundation models Light-O1, Light REACT and LightNav-0
BeingBeyond 北京智在无界科技有限公司
China
Beijing embodied-AI startup building general-purpose humanoid robot foundation models trained on large-scale human data, plus dexterous-hand hardware.
Build AI
United States
San Francisco data company collecting egocentric video of real factory workers; its open Egocentric-100K set holds 100,405 hours for robot learning.
Ropedia
Singapore
Singapore physical-AI data company founded 2025; its HOMIE head-mounted rig and Xperience datasets supply human demonstration data to embodied-AI builders.
Swarmanoid Project
Belgium
EU FP7 research project (2006-2010) led by IRIDIA at Universite Libre de Bruxelles; designed heterogeneous foot-bot, hand-bot, and eye-bot swarm robot systems.
Xi'an Blue Worm EAI Technology Co., Ltd. 西安蓝虫具身智能科技有限公司
China
Xi'an embodied-AI startup founded in 2025 by a Xi'an Jiaotong University team, making the modular Mantis Standard wheeled dual-arm robot with VR teleoperation.
Songying Technology 松应科技
China
Songying Technology (ORCA3D) is a Shanghai physics-AI simulation software maker for embodied-AI robot training, backed by early-stage VC investors.
Config
United States
Builds data infrastructure and foundation models for general-purpose bimanual robotic manipulation.
MediaMeta
Singapore
Singapore's MediaMeta runs a nine-product AI platform built on human-behavioral-modeling and social-intelligence technology for content and marketing.
Robotics needs a service framework.
RSF defines a common language for robot service capability, lifecycle operations, certification pathways, and service-provider networks.
Join the RobotToday community on LinkedIn
Daily robotics news, in-depth analysis, conference highlights, and discussions with professionals worldwide.
