ShenDuZhiYao Secures Nearly $50 Million in Series D Funding for AI-Enabled Full-Package Clinical Trials | Exclusive from 36Kr

Text by | Hu Xiangyun

Edited by | Hai Ruojing

36Kr learned that Deepwise, a global leader in AI-driven drug discovery, recently secured nearly $50 million in Series D financing. This round of financing was led by CDH Baifu, with existing shareholders New Summit Capital and Sequoia China continuing to increase their investments. Exponent Capital served as the exclusive financial advisor. The raised funds will be mainly used for the technological R & D iteration of the "Multi-Agent Collaboration Network" and the construction of the global delivery network.

Deepwise was founded in 2017. Compared with its early exploration of single technology points, in the past three years, Deepwise has completed a generational leap from "single-point AI technology verification" to an "AI-Native Clinical Research Platform". This evolution has enabled it to break away from the category of traditional software providers and transform into a core business partner capable of delivering full-process results of clinical trials.

In the view of Li Xing, the founder and CEO of the company, the future of pharmaceutical R & D lies not in the replacement of single functions but in the reconstruction of cognition. With the explosion of generative AI, Deepwise was the first to upgrade its underlying NLP capabilities to a "Multi-Agent Collaboration System" with tens of thousands of vertical domain agents.

"We no longer deliver single-function modules but an 'AI Agent Cluster' capable of collaboratively completing the full process of clinical trials," Li Xing told 36Kr. "Our core competitive advantage lies in using the 'Cognitive Atomism' to reconstruct the R & D process. The system breaks down complex clinical trials into tens of thousands of tiny atomized tasks, each of which is handled by a specialized agent. These agents are connected through a synaptic network similar to the brain nerves, achieving a level of professionalism far beyond that of general large models."

Image source: Deepwise

To some extent, the evolution of this technological architecture has promoted the transformation of the business model. While the industry generally adopts the traditional "pay-by-headcount/hour" model, Deepwise has begun to explore the Outcome-based Model.

The ability to achieve this mainly stems from the mutual verification mechanism between the "Planning Agent" and the "Execution Agent" built by Deepwise. When the AI is corrected by human experts in actual projects, the system will trigger a "Self-Reflection Mechanism" to automatically trace back and correct the code logic. This unique "Two-Way Verification Feedback Flywheel" mechanism endows the system with an intuition similar to that of human experts, enabling it to draw inferences from one instance.

Taking the cooperation between the company and Japanese innovative pharmaceutical company Immunorock as an example, Deepwise's service model is equivalent to providing it with a "Digital Rehearsal". It is reported that in traditional clinical trial protocol design, it often relies on the personal experience of experts, which is prone to logical loopholes. However, through the "Digital Twin" technology, Deepwise's system conducts a full-process simulation before actual patient enrollment, simulating the entire process from patient screening to data statistics and avoiding potential risk points that could lead to a high dropout rate in advance. Eventually, this protocol helped the client obtain a one-time approval from the Pharmaceuticals and Medical Devices Agency (PMDA) in Japan.

In terms of data security, regarding the most sensitive issue of data sovereignty for pharmaceutical companies, Deepwise adheres to the principle of "data does not land, and the model does not memorize". Each client project runs in an independent physical sandbox, which is destroyed after the project ends. Only desensitized "error logics" are retained to improve the robustness of the system.

According to data provided by Deepwise, the company has served more than 1,000 pharmaceutical companies in total, and through the actual delivery of more than 40,000 projects, the universality and stability of this system in complex pharmaceutical scenarios have been verified.

"The pharmaceutical R & D industry is at a historical juncture of transformation from 'labor-intensive' to 'intelligence-intensive'," Li Xing said. "Deepwise's vision is not to be a service provider but to build a new generation of 'Pharmaceutical R & D Operating System'. Through AI agents, it will automate cumbersome processes, freeing scientists from data documents and allowing them to focus on real scientific innovation."