Whale Leap Dynamics Secures Multi-Million Yuan Seed Round Investment from Xinghaitu, Creating Ready-to-Use Robo Labor with "Data + Model + End Execution" | Emerging New Project

Text by | Wang Xinyi

Edited by | Qiu Xiaofen

One-sentence Introduction

Whale Leap Dynamics was founded in 2026. With the closed-loop of "data + model + end execution", it provides Robo Labor (robot labor force) that is scenario-oriented and can be quickly deployed. It enables physical labor force to be subscribable, elastically expandable, and ready-to-use, just like AWS computing power, and can replace humans in engaging in physical operations that are dangerous, arduous, dirty, and repetitive (Deadly/Difficult/Dirty/Duplicate).

Financing Progress

Recently, Whale Leap Dynamics has completed a seed round financing of tens of millions of yuan led exclusively by Xinghai Map. Shendu Capital served as the exclusive financial advisor.

This round of financing will be used for team growth, mass production and delivery of products, R & D of expert skills, and work related to data collection and operation.

Xinghai Map said: "In the second half of embodied intelligence, it's all about the systematic ability of 'data efficiency × model generalization × end execution'. Whale Leap Dynamics focuses on the data main line and is dedicated to achieving rapid delivery and iteration in real scenarios, which is highly consistent with our concept of pursuing a closed-loop of data in the physical world. The team has the scarce genes of data, model, and delivery. Such a team combination is rare in the industry. Therefore, we are optimistic that it will become a global leading pioneer in embodied intelligence applications starting from high-quality data."

Products and Business

Whale Leap Dynamics targets the To B market. Focusing on data, it is dedicated to achieving rapid delivery and iteration in real scenarios and provides a software-hardware integrated solution of "data + model + end actuator".

Overseas, the practices of companies such as Sunday, Generalist AI, and Genesis AI have verified that data is the first principle for the large-scale implementation of embodied intelligence, and have converged on the data solution of "Ego-centric + UMI", that is, "first-person perspective + universal operation interface".

Li Guangyu, the founder of Whale Leap Dynamics, believes that it is unnecessary to overly pursue the complexity of super-large models. As long as a closed-loop is formed among real data, scenario-based models, and end execution, engineering delivery and large-scale implementation can be quickly achieved.

Therefore, Whale Leap Dynamics adopts the "data collection ability + model ability + in-depth cognition of dexterous operation" model to achieve the implementation of embodied intelligence in scenarios.

The core barriers of Whale Leap Dynamics are mainly reflected in the following three aspects:

1. Self-developed Ego-centric + UMI data collection system: It can achieve sub-millimeter pose positioning and sub-millisecond multi-source time synchronization, cover visual, force, pose, and environmental interaction signals, and conduct real-time scene understanding and analysis at the edge side to ensure data quality.

2. One million hours of data pipeline and human-in-the-loop strategy: A data pipeline is established, with the ability to clean, annotate, and analyze data at the one million-hour level. Through large-scale real data to cover complex working conditions, and at the same time introduce the Human-in-the-Loop paradigm. With the immediate error correction ability of humans, it ensures that the robot can continuously evolve on the premise of being available from the first day.

3. Self-developed 3D world model and expert skills: It includes the ability of in-depth spatial understanding and precise mobile operation, and a high-fidelity physical cognitive engine is built on the data base. Based on this model, the robot can not only "identify objects", but also understand gravity, friction, deformation trends, and interaction boundaries, truly realizing the paradigm shift from "perception - execution" to "cognition - prediction - adaptation".

△ Visualization of data collection by Whale Leap Dynamics. Image source: Provided by the enterprise

It is reported that the business model of Whale Leap Dynamics has also been recognized by customers and industry players. Taking the application scenario of material handling as an example, customers can seamlessly integrate robots into the existing business process, achieving the abilities of autonomous movement in indoor and outdoor full scenarios, high-precision handling, flexible loading and unloading, and standard carrier operation, effectively solving the pain points of the industry.

Currently, Whale Leap Dynamics has cooperated with many leading enterprises in the manufacturing and logistics industries to promote the implementation of its products in multiple scenarios.

Team Introduction

Dr. Li Guangyu, the CEO of Whale Leap Dynamics, graduated from the Department of Electrical Engineering at the University of Southern California (USC). He served as the person in charge of embodied data and dexterous operation at the Beijing Humanoid Robot Innovation Center. He built a dexterous operation and embodied data team from scratch, led benchmark projects such as the RoboMIND multi-modal embodied data set and the embodied intelligent agent "Huisi Kaiwu", and has rich experience in data and model. Previously, he led the construction of a one-million-level autonomous driving data closed-loop and simulation system at Didi and Qingzhou Zhihang, and has dual industrial practical experience in embodied intelligence and autonomous driving.

△ Dr. Li Guangyu, the founder and CEO of Whale Leap Dynamics. Image source: Provided by the enterprise

The core team covers the entire chain of data, model, hardware, software, and global business. It is composed of core backbones from leading enterprises such as Geek+ Robotics, Neolix, Qingzhou Zhihang, Beijing Humanoid, Li Auto, Xiaomi, and Didi. Team members graduated from top universities such as USC, Tsinghua University, Zhejiang University, and the University of Science and Technology of China. They have in-depth knowledge in the fields of robotics, reinforcement learning, and multi-modal large models, and also have champions of top competitions such as RoboMaster.

The team has rich experience in overseas R & D, compliance certification, and local deployment, and has the complete implementation ability from cutting-edge R & D to mass production of ten thousand units and delivery of one million-level data.

Founder's Thoughts

• A company must become the optimal solution in a niche area to win the competition in the long run.

The cruelty of the To B market lies in that having a working robot is not enough, and having a robot with a positive unit economic model (UE) is also not enough. It must become the optimal solution in a niche area to win the competition in the long run.

The key to the real implementation of products lies in that the enterprise has a deep understanding of the existing workflow, can seamlessly integrate the product into the business chain, and target the links with the highest labor consumption and the most obvious ROI. The hardware design needs to be practical, and the data collection should be bold and aggressive. Only in this way can the closed-loop of scenario running, value verification, data feedback, and product iteration be gradually achieved.

• Scenario-based robot companies will be the winners in the future.

Currently, the industry barrier is shifting from single-point algorithms to systematic capabilities. The core competitive points of enterprises are: first, the "data cost of unit generalization ability", and second, the "density of large-scale deployment in scenarios".

Data efficiency determines whether we can generalize the operation ability of robots to new materials, new carriers, and new sites at a very low marginal cost; the implementation density determines whether a scale effect can be formed within a single scenario.

The future winners will be those "scenario-based robot companies" that can use the most accurate data to leverage the maximum generalization boundary and occupy the core scenarios with the highest density.

• Standardized product companies in the robot field are emerging.

Currently, the hardware supply chain is becoming increasingly mature, the data infrastructure is being scaled up, and the capabilities of large models and agents are booming. The new AI-driven paradigm is expected to enable robot companies to shift from the traditional "project customization and integration" to "standardized product companies".

It enables robot enterprises to define hardware with a product-oriented mindset and deliver value with the logic of agents. By achieving the ultimate in intelligence in vertical scenarios, combined with standardized hardware and clear product definitions, they are fully capable of achieving high-explosive and large-scale commercial shipments.

• The ecological niche that Whale Leap Dynamics wants is: data + application.

The global industry's demand for flexible, low-cost, and highly reliable physical labor force has never declined.

Embodied intelligence will repeat the evolution trajectory of autonomous driving, from the early attempts of Robotaxi (driverless taxis), to the large-scale mass production of assisted driving, and then to the extension of vertical scenarios such as unmanned delivery, trunk logistics, and mining trucks. For embodied intelligence, it will also follow the path from the verification period, engineering convergence period to the commercial explosion period.

Whale Leap Dynamics is building an ecologically friendly and general skill platform. We are good at the links of data-driven, closed-loop, and implementation and deployment. We pay more attention to the combined value of "end actuator + skills". We hope that after customers connect the hardware, connect the system and sensors, they can directly adjust the skills to meet their needs.

Cover source | Provided by the enterprise

Welcome to communicate!

Scan the code to join the "Intelligent AI Communication Group"