MindRank Secures $52M Series B Funding as AI-Designed Weight-Loss Drug Enters Phase III Trials

Text | Hu Xiangyun

Edited by Hai Ruojing

36Kr has learned that MindRank recently completed a $52 million Series B financing round, with investors including top-tier RMB and USD funds. Caic Capital served as the exclusive financial advisor. The raised funds will be used to upgrade and iterate its AI drug discovery engine, the Molecule Arts Platform (MAP), enhance its Multi-Agent collaborative system and Clinical Data-in-the-Loop framework, advance the Phase III clinical trials and commercialization of its self-developed oral small-molecule GLP-1 candidate MDR-001, and expand its pipeline of other differentiated innovative drugs.

MindRank’s vision is to become an AI-native biopharmaceutical company. Its founder and CEO, Niu Zhangming, brings years of experience as the CTO of a publicly listed AI healthcare company in Germany. He believes that traditional new drug development has heavily relied on the expertise of senior scientists, requiring the synthesis of hundreds to thousands of molecules through iterative trial-and-error processes. However, breakthroughs in AI technology are now poised to continuously disrupt this conventional approach, transforming the “trial-and-error” model—akin to finding a needle in a haystack—into a data-driven, computable, and iterable paradigm of “standardized creation.”

This is also the original intention behind MindRank's establishment of MAP.

Schematic of MindRank’s Molecule Arts Platform (MAP) Technology Platform (Image source: MindRank)

According to the introduction, the Molecule Arts Platform (hereinafter referred to as “MAP”) features a three-tier architecture: The first tier is the AI design layer, which integrates the company’s self-developed PharmkGPT biology platform, the Molecule Pro small-molecule drug discovery platform, and the Molecule Dance structural biology platform. This layer covers the entire workflow for various types of drugs, including small molecules and peptides, from preclinical development to Investigational New Drug (IND) application.

The second-layer dry-wet complementary experimental platform primarily serves as a data supply hub, driven by the synchronized operation of the Proxima Matrix dry-lab simulation system and the Proxima Foundry wet lab, continuously generating high-quality training data to support AI model iteration. “MindRank independently trains its foundational large model for biopharmaceuticals using 200 GPUs and has built its own data production pipeline. The goal is to enable pharmaceutical companies to evolve into AI-native enterprises capable of rapid iteration, while delivering customized, high-quality data required by models through standardized laboratory processes,” stated Niu Zhangming.

The third layer is the Clinical Pro clinical AI platform. Rather than simply accumulating clinical data, it integrates clinical feedback into the R&D system to refine earlier computational and experimental judgments. This constitutes MindRank’s core competitive barrier distinguishing it from other AI-driven pharmaceutical companies. Part of this feedback stems from clinical research data or clinical data accumulated through MindRank’s proprietary clinical pipeline; for instance, Clinical Pro has already amassed over 1,300 de-identified, clinically relevant data points from its GLP-1 pipeline candidate, MDR-001.

“We do not define MAP merely as an AI tool, but rather as a drug discovery system capable of continuous learning and evolution. Its core vision is toBy leveraging an AI multi-agent assistant to connect end-to-end data from preclinical to clinical stages, the ‘Clinical Data-in-the-Loop’ approach enables continuous self-iteration. In this way, real-world data generated by each pipeline, regardless of success or failure, can feed back to enhance platform capabilities, thereby reducing the cost and failure rate of subsequent new drug development..” Niu Zhangming stated.

Currently, MAP’s first validated drug candidate, MDR-001, entered Phase III clinical trials in late 2025. The company has completed enrollment of 760 subjects in China, collaborating with nearly 50 clinical trial centers. If all goes well, the drug is “expected to be launched in China within the next two to three years,” positioning it to compete in the GLP-1 market valued at hundreds of billions of US dollars.

Reviewing the development journey of MDR-001, the MindRank team confirmed the preclinical candidate (PCC) within eight months by synthesizing more than 80 novel small molecules. The entire process, from project initiation to the launch of Phase III clinical trials, took approximately four and a half years and cost around $23 million. According to data provided by MindRank, this efficiency is more than ten times the industry average. In contrast, under traditional methods, developing an innovative drug from design to entry into Phase III clinical trials typically requires seven to nine years and costs $300–400 million.

This high efficiency is not an isolated case. Empowered by AI technology, MindRank’s team of just over 40 employees has, since its inception, advanced 15 innovative pipelines, established a portfolio of five preclinical candidates (PCCs), and secured three Investigational New Drug (IND) approvals in both China and the United States. Following MDR-001, MRANK-106, a dual WEE1 & YES1 inhibitor targeting solid tumors with limited treatment options such as pancreatic cancer, has also received clinical trial approval. In terms of drug modalities, MindRank’s pipeline assets encompass various cutting-edge approaches, including GPCR-targeting agents, molecular glues, allosteric inhibitors, and dual-target small-molecule oncology drugs.

Zhangming Niu believes that as AI technology gradually permeates all aspects of the company’s R&D system, the value of future AI-native pharmaceutical companies will no longer be confined within the traditional biotech framework. “The valuation logic for traditional biotechs stems from a single-drug Net Present Value (NPV) model, whereas AI-native pharmaceutical enterprises validate more efficient methods for drug production. In the case of MindRank, this is embodied by the MAP system. As an increasing number of drug pipelines advance, MAP can transfer prior R&D experience to subsequent pipelines, creating a continuously reusable R&D flywheel.”

This was also the path followed by the previous generation of CADD (Computer-Aided Drug Design) companies. Take Vertex Pharmaceuticals, a global CADD pioneer with a market capitalization of $120 billion, as an example. By employing a strategy of precise computer-aided molecular design, the company has continuously launched several blockbuster drugs with annual sales exceeding $10 billion each, growing into a pharmaceutical giant with a valuation in the hundreds of billions. Its experience demonstrates that mastering the most advanced computational tools and achieving the highest prediction accuracy may enable leveraging greater commercial returns with more streamlined resources.

Today, MindRank aims to become the Vertex Pharmaceuticals of the AI era, ensuring that “every molecular design, experimental dataset, and clinical outcome serves as the starting point for the next round of innovation.” The company is committed to continuously advancing blockbuster drug candidates into clinical trials and ultimately commercialization. Revenue generated from drug sales will provide a stable cash flow to fuel the iterative development of its AI technology platform, thereby consistently producing novel drug assets with global competitiveness.