CataX AI
AI-native biotech using AI agents to accelerate early-stage clinical drug development.
Website: https://www.catax.bio/
Cover Block
From the public record
| Field | Value |
|---|---|
| Name | CataX AI |
| Tagline | AI-native biotech using AI agents to accelerate early-stage clinical drug development. [TechNode, November 2026] |
| Headquarters | Shanghai, China [TechNode, November 2026] |
| Founded | 2026 [TechNode, November 2026] |
| Stage | Seed [TechNode, November 2026] |
| Business Model | Other |
| Industry | Deeptech |
| Technology | AI / Machine Learning |
| Geography | East Asia |
| Growth Profile | Venture Scale |
| Founding Team | Co-Founders (2), Wang Zilong and Jiang Zhemin [TechNode, November 2026] |
| Funding Label | Seed |
| Total Disclosed Funding | RMB 40 million, about $5.5 million [TechNode, November 2026] |
Links
From the public record
- Website: https://www.catax.bio/
- Career: https://catax.app/career
The Short Version
PUBLIC CataX AI is a Shanghai-based AI-native biotech aiming to move drug candidates through early clinical development with AI agents, and it merits attention now because it raised RMB 40 million in seed funding in November 2026 from Qiming Venture Partners, which also reportedly incubated the company [TechNode, November 2026] [36Kr, November 2026] [Sohu, November 2026]. Founded in 2026, the company is entering the market with a hybrid model rather than a pure software pitch: its website says it sources drug assets, develops clinical proof of concept, and advances programs with partners, while outside coverage describes a plan to license or co-develop preclinical or early-clinical assets and push them toward proof of concept [CataX AI, retrieved 2026] [TechNode, November 2026].
The point of differentiation, if the model works, is operational structure more than model novelty. CataX frames this as a "Clinical-of-One" approach in which a clinical lead works with AI agents on research review, data analysis, and document preparation, with human experts retaining key decisions, according to company materials and summarized external reporting [CataX AI, retrieved 2026] [TechNode, November 2026].
The founding bench is one of the more investable parts of the early story, though the public record is still thin. TechNode describes founder and CEO Wang Zilong as having a background in clinical medicine and AI research at Microsoft Research Asia, while Microsoft Research lists his M.D. from Fudan University and research interests in AI for healthcare; co-founder Jiang Zhemin is reported to bring more than 15 years of clinical-development experience, with prior roles at Tigermed and BeiGene according to 36Kr [TechNode, November 2026] [Microsoft Research, retrieved 2026] [36Kr, November 2026].
The financing picture is simple at this stage. Multiple outlets reported the same November 2026 seed round at RMB 40 million, or about $5.5 million at the reported converted figure, with Qiming as sole investor, and no public source in the provided record discloses valuation or additional backers [TechNode, November 2026] [凤凰网, November 2026] [动点科技, November 2026].
Over the next 12 to 18 months, the key question is less whether CataX can describe an agentic workflow and more whether it can show repeatable evidence that this workflow shortens timelines or improves decision quality in asset selection and early clinical execution. Public reporting has not yet surfaced named programs, counterparties, or development outcomes, so the next proof points are likely to be program disclosures, partnership evidence, and signs that the company can translate founder-market fit into a durable operating engine [TechNode, November 2026] [CataX AI, retrieved 2026].
Single-source, plausible -- Core funding is corroborated by multiple public reports, but product details and operating model rely partly on company materials, and team background details are only partially corroborated by independent sources.
Taxonomy Snapshot
| Axis | Value |
|---|---|
| Stage | Seed |
| Business Model | Other |
| Industry / Vertical | Deeptech |
| Technology Type | AI / Machine Learning |
| Geography | East Asia |
| Growth Profile | Venture Scale |
| Founding Team | Co-Founders (2) |
| Funding | Seed, total disclosed about $5.5 million [TechNode, November 2026] |
The Company in Brief
PUBLIC
CataX AI surfaced publicly in 2026 as a Shanghai-based company trying to compress one of biotech’s slower workflows: early clinical development. The company presents itself as an "AI-native biotech" rather than a pure software vendor, and its own description is unusually direct about the model, saying it sources drug assets, develops clinical proof of concept, and advances programs with partners [CataX AI]. Public-facing company materials also identify CataX AI as the brand of Shanghai CataX Biopharmaceutical Technology Co., Ltd. [CataX AI].
The chronology that can be supported from company-controlled sources is still short, which matters for interpretation. The website ties the business to Shanghai and shows a product and corporate identity already in market under the CataX AI name in 2026, while the About and Approach pages name the underlying legal entity in English and Chinese [CataX AI]. That leaves the current public record at an early stage: a newly formed biotech platform with a defined operating concept, a stated intent to move partner assets toward clinical proof of concept, and a legal entity that is at least publicly disclosed on its own site [CataX AI].
Unconfirmed -- This section relies primarily on company website disclosures, with legal-entity and operating-description support from CataX AI's own public pages [CataX AI].
What They Have Built
MIXED
The product claim is straightforward, even if the operating details remain thin in public. CataX AI describes itself as an "AI-native biotech" that sources drug assets, develops clinical proof of concept, and advances programs with partners, framing the company less as a pure software vendor and more as a development platform around early-stage drug programs [CataX AI, retrieved 2026]. TechNode's reporting aligns with that positioning, stating that the company plans to license or co-develop drug candidates at the preclinical or early-clinical stage and move them toward clinical proof of concept [TechNode, November 2026].
The more differentiated element is the operating model the company calls "Clinical-of-One." On CataX AI's website, that model is presented as one full-stack clinical lead working with an agent-powered workbench and external experts across functions including medical, regulatory affairs, pharmacovigilance, clinical operations, data management, and statistics [CataX AI, retrieved 2026]. Separate public research summaries describe the same concept more explicitly: a clinical lead directs AI agents to review research, analyze data, and prepare documents, while human experts retain decision-making authority [TechNode, November 2026]. That distinction matters because it suggests the company is applying AI to compress clinical-development workflows rather than claiming autonomous drug-development execution.
What cannot yet be verified from public evidence is the underlying technical stack, model performance, or any demonstrated proof that the system improves development outcomes against standard CRO-led processes. The website states that CataX AI "improves itself" and that humans and agents "work as one," but those remain company-level claims rather than independently validated product benchmarks [CataX AI, retrieved 2026]. There is also no verified public demo, named customer deployment, or disclosed program outcome in the available sources, so the present read is that the technology narrative is conceptually coherent but still early in disclosed evidence [36Kr, November 2026].
Unconfirmed -- This section relies materially on company website claims, with partial support from TechNode and 36Kr on the operating model and business scope.
Market Size and Demand
PUBLIC
This market matters now because drug discovery output has expanded faster than the industry’s ability to turn candidates into clinical proof of concept, and CataX AI is positioned in that bottleneck rather than at the earlier screening layer [CataX AI, retrieved 2026].
The public record does not support a conventional TAM, SAM, and SOM build for CataX AI itself. What is available is narrower but still useful: the company frames its addressable problem around the gap between a large preclinical asset base and a smaller flow of programs entering clinical development, citing 12,704 preclinical drugs worldwide as of January 2025 and 1,868 NMEs entering clinical development during 2025 [CataX AI, retrieved 2026]. Those figures are analogous market indicators rather than a revenue market size, but they do point to a large inventory of assets competing for development bandwidth, capital, and operational execution.
| Market signal | Figure | Scope | Source context |
|---|---|---|---|
| Preclinical drugs worldwide | 12,704 | Global, January 2025 | Cited by CataX AI from Citeline via Vizient [CataX AI, retrieved 2026] |
| NMEs entering clinical development | 1,868 | Global, during 2025 | Cited by CataX AI from DXY / BCG [CataX AI, retrieved 2026] |
The gap between these two figures does not establish attrition economics on its own, but it does support the basic premise that there is a large upstream supply of drug assets relative to the annual flow into human studies [CataX AI, retrieved 2026]. For an operator trying to build around early clinical development, that is the relevant market signal: too many candidates, not enough efficient progression.
Demand drivers follow from that mismatch. CataX AI’s own materials argue that "drug candidates are piling up" while approvals remain broadly flat, and its operating model is built around compressing the work of research review, data analysis, and document preparation under a clinically led workflow [CataX AI, retrieved 2026]. TechNode’s reporting is directionally consistent with that framing, describing the company as using AI agents to support early-stage clinical drug development and planning to license or co-develop preclinical or early-clinical candidates through proof of concept [TechNode, November 2026]. The demand case, then, is less about software seat expansion and more about whether sponsors and asset owners increasingly want development partners that combine clinical judgment with automation.
The adjacent markets are clear even if the boundaries are still soft. One adjacency is AI-enabled biopharma R&D, where vendors and drug developers use machine learning to improve target discovery, trial design, and development workflows; another is contract research and clinical development services, where execution capacity and regulatory process matter more than algorithmic novelty. CataX AI appears to sit between those categories, with a model that resembles an asset-centric biotech combined with an AI-assisted development stack rather than a pure software platform [CataX AI, retrieved 2026] [TechNode, November 2026]. That positioning could widen the opportunity set if it works, but it also means the relevant comparables may come from biotech, techbio, and clinical services rather than one clean market bucket.
Regulatory and macro forces also matter because early clinical development is documentation-heavy, closely supervised, and expensive to get wrong. The company’s public materials emphasize human oversight, with agents supporting work while key decisions remain with people, which is a sensible positioning choice in a regulated environment where accountability cannot be delegated to tooling alone [CataX AI, retrieved 2026]. At the same time, the broader funding backdrop appears constructive enough for specialist AI-biotech formation in China, evidenced here by Qiming Venture Partners both incubating and solely backing the seed round, although one financing event should be read as sponsor conviction rather than proof of broad market adoption [TechNode, November 2026] [36Kr, November 2026].
Inferred, not confirmed -- Market sizing in this section relies on company-cited analogous indicators rather than independent third-party market reports, with partial support from TechNode on the clinical-development use case.
Who Else Is Fighting for This
MIXED CataX AI is positioning itself less like a software vendor and more like an AI-enabled development operator, which places it in a different lane from most life-science tooling companies and closer to early-stage biotech platforms that combine capital, scientific judgment, and execution capacity [CataX AI, retrieved 2026] [TechNode, November 2026].
The competitive map is easiest to read in three layers. First are incumbent contract research and clinical-development service firms, the sort of organizations that provide the regulatory, data-management, pharmacovigilance, and trial-operations work that CataX AI's website depicts inside its "Clinical-of-One" workflow [CataX AI, retrieved 2026]. Second are biotech developers and asset-centric platforms that license or co-develop programs and try to carry them toward inflection points, which is closer to the operating model described in TechNode's reporting on CataX AI's plan to advance preclinical or early-clinical assets to clinical proof of concept [TechNode, November 2026]. Third are adjacent AI tooling groups, including internal teams at pharma companies and research organizations, that may not compete for the same economics but do compete for the same promise: compressing the labor and cycle time of development work through software and automation [CataX AI, retrieved 2026]. In practice, that means CataX AI is not only competing against startups. It is also competing against the default decision to keep development inside a CRO-led process or inside a pharma partner's own organization.
Where the company appears to have an edge today is in founder mix and sponsor alignment, although both advantages look early rather than entrenched. TechNode and 36Kr describe Qiming Venture Partners as the sole seed investor, and TechNode reports that Qiming also incubated the company, which matters because this model likely needs access to assets, scientific networks, and follow-on capital at the same time [TechNode, November 2026] [36Kr, November 2026]. On the team side, public materials point to Wang Zilong's clinical-medicine training and AI research background at Microsoft Research, while 36Kr ties co-founder Jiang Zhemin to prior roles at Tigermed and BeiGene [Microsoft Research, retrieved 2026] [36Kr, November 2026]. That combination is relevant to the company's claim that a clinical lead can direct AI agents across research, analysis, and document preparation, but it is still perishable. Talent is durable only if it turns into a repeatable operating system, and no public source yet shows named programs, partners, or clinical outcomes that would prove the system scales beyond the founding team [TechNode, November 2026] [CataX AI, retrieved 2026].
The exposure is clearer than the moat. CataX AI does not appear to own a distribution channel, a public customer base, or a disclosed portfolio of drug assets, which leaves it vulnerable to better-capitalized substitutes on both sides of the market [TechNode, November 2026]. If an asset owner wants speed with established execution capacity, organizations with deep clinical-operations benches such as Tigermed, where Jiang previously worked according to 36Kr, hold an institutional advantage in process, staffing, and existing sponsor relationships [36Kr, November 2026]. If a large pharma company wants AI assistance without giving up program control, internal development teams and incumbent external vendors can absorb similar agentic tooling into their existing workflows. That does not invalidate CataX AI's model, but it does mean the company must prove it is better at selecting assets and converting development work into proof-of-concept value than the service stack it is trying to compress.
The most plausible 18-month scenario is a sorting of models rather than a winner-take-all outcome. Qiming Venture Partners looks like the most likely winner if incubation plus concentrated capital can help CataX AI secure attractive early assets and demonstrate one or two visible development milestones, because the seed round gives the firm a direct seat in testing whether this hybrid model can produce proof-of-concept efficiently [TechNode, November 2026] [Sohu, November 2026]. The most likely loser if that does not happen is the broad category of independent, stand-alone AI-for-biotech tooling efforts that lack both asset ownership and clinical execution depth, since CataX AI's thesis implicitly argues that software alone is not enough [CataX AI, retrieved 2026] [TechNode, November 2026]. The unresolved point is whether CataX AI itself can turn that thesis into evidence before incumbents and in-house teams incorporate similar agent workflows into better-established development machines.
Single-source, plausible -- Core positioning and financing are corroborated by TechNode and 36Kr, but competitive mapping and substitute analysis rely partly on company materials and inference from the disclosed operating model.
Opportunity
PUBLIC
The prize here is unusually large because CataX AI is not trying to sell a narrow workflow tool, it is trying to compress the slowest part of biotech value creation, moving early assets toward clinical proof of concept, into a faster, more repeatable operating system for drug development [TechNode, November 2026] [CataX AI, retrieved 2026].
The headline opportunity is to become an AI-native development platform that repeatedly takes preclinical or early-clinical assets through the first meaningful value inflection point, then monetizes through licensing, co-development, or downstream partnering [TechNode, November 2026] [CataX AI, retrieved 2026]. That outcome is still early, but it is not abstract. The company has been explicit that it intends to source drug assets, develop clinical proof of concept, and advance programs with partners, while its operating model centers on a single clinical lead directing agent-based research, analysis, and document work [CataX AI, retrieved 2026] [TechNode, November 2026]. If that model works in practice, the upside is closer to a scaled biotech engine than to a software vendor with service-heavy margins.
The early public evidence matters because it points to alignment between product concept, team formation, and capital source. Qiming Venture Partners is reported as both incubator and sole seed investor in the RMB 40 million round announced on November 16, 2026, which suggests a tighter build plan than a typical first financing [TechNode, November 2026] [36Kr, November 2026]. On team, Wang Zilong's public profile ties together clinical medicine training and AI research at Microsoft Research, while Jiang Zhemin is reported to bring more than 15 years of clinical-development experience, including roles at Tigermed and BeiGene [Microsoft Research, retrieved 2026] [TechNode, November 2026] [36Kr, November 2026]. That does not prove execution, but it does make the central bet, clinical judgment paired with agentic automation, more reachable than a pure concept-stage claim.
| Scenario | What happens | Catalyst | Why it's plausible |
|---|---|---|---|
| Proof-of-concept factory | CataX AI builds a repeatable model for advancing multiple in-licensed or co-developed assets to clinical proof of concept, then earns economics through partnering or out-licensing | The first program reaches a visible proof point that validates the "Clinical-of-One" workflow and partner model [CataX AI, retrieved 2026] | The company is explicitly organized around sourcing assets and advancing them with partners, not merely selling software seats [CataX AI, retrieved 2026] [TechNode, November 2026] |
| Agentic backbone for small biotech | CataX AI packages its internal workbench and operating method into a platform used by external biotech teams for research, data analysis, and document preparation | A first external development partnership or structured co-development agreement demonstrates that the platform can travel beyond internal use [TechNode, November 2026] | Public materials already describe an agent-powered workbench and a division of labor between clinical leads, external experts, and agents, which is consistent with a productizable workflow layer [CataX AI, retrieved 2026] |
| China-to-global development bridge | CataX AI becomes a preferred early-development partner for under-resourced programs in China that need faster preparation for broader partnering | Qiming's network and incubation role help the company aggregate overlooked assets and counterparties into one funnel [TechNode, November 2026] [Dealroom News] | The company is Shanghai-based, focused on early development, and backed solely by a top-tier healthcare investor that reportedly incubated the business, which can matter in sourcing as much as in financing [TechNode, November 2026] [Dealroom News] |
What compounding could look like is straightforward, even if it is not yet proven publicly. Each asset worked through the system should generate process knowledge in protocol design, evidence review, data interpretation, and regulatory documentation, all of which can improve future cycle times if captured correctly inside the agentic workbench [CataX AI, retrieved 2026]. The company itself frames this as an AI-native biotech that "improves itself," and its site architecture describes a full-stack clinical lead supported by specialized functions and agents rather than a loose collection of tools [CataX AI, retrieved 2026]. The important implication is that one successful program may do more than produce one asset outcome, it may also refine the internal machine that handles the next asset.
That flywheel would be stronger if two conditions hold. First, better asset selection and faster development work would need to raise the hit rate on which programs are worth advancing. Second, the resulting evidence package would need to make CataX AI a more credible counterparty for future licensors and co-development partners [TechNode, November 2026] [CataX AI, retrieved 2026]. Publicly, there is no named program, customer, or partnership yet, so this remains an informed scenario rather than an observed operating pattern. Still, the structure of the model is one where throughput and learning can compound if early execution is real.
The size of the win is best framed by the value of reaching clinical proof of concept in a market where upstream asset supply is already large. CataX AI's own site cites 12,704 preclinical drugs worldwide as of January 2025, attributed to Citeline via Vizient, against 1,868 new molecular entities entering clinical development during 2025, attributed to DXY and BCG [CataX AI, retrieved 2026]. Even allowing for marketing framing, the directional point is credible: there are more candidate assets than the industry can move forward efficiently. If CataX AI became a trusted engine for selecting and advancing even a small slice of that backlog, the company could plausibly support a venture-scale outcome through a mix of asset economics and platform value. No public valuation benchmark is available for CataX AI today, and no direct public comparable in the provided source set supports a hard range. The upside case is therefore best stated qualitatively: if the proof-of-concept factory scenario works, this could become a high-value biotech platform rather than a single-asset company (scenario, not a forecast) [TechNode, November 2026] [CataX AI, retrieved 2026].
Single-source, plausible -- Core financing and strategy are corroborated by TechNode and 36Kr, but the operating model and market framing rely materially on company materials.
Sources
From the public record
[TechNode, November 2026] CataX AI raises RMB 40 million to bring AI agents into clinical drug development · TechNode | https://technode.com/2026/11/16/catax-ai-raises-rmb-40-million-to-bring-ai-agents-into-clinical-drug-development/
[36Kr, November 2026] Qiming Venture Partners Provided Him with His First Round of Startup Funding | https://eu.36kr.com/en/p/4071739103187846
[CataX AI, retrieved 2026] CataX AI | https://www.catax.bio/
[Microsoft Research, retrieved 2026] Zilong Wang at Microsoft Research | https://www.microsoft.com/en-us/research/people/wangzilong/
[Dealroom News, retrieved 2026] Qiming backs AI-native biotech CataX AI in sole seed bet | Dealroom News | https://dealroom.co/news/172821-qiming-backs-ai-native-biotech-catax-ai-in-sole-seed-bet/
[Reuters, March 2015] China's CICC expects M&A deals jump on state sector reforms | https://www.reuters.com/article/cicc-ma-idUSL4N0WF20F20150316/
Articles about CataX AI
- CataX AI’s Clinical-of-One Model Directs AI Agents for Early-Stage Drug Development — The Shanghai startup, incubated by Qiming Venture Partners, plans to license or co-develop drug candidates and advance them to clinical proof of concept.