BioGeometry (百奥几何)
Beijing generative-AI protein design company led by Mila professor Tang Jian, licensing GeoFlow antibody and enzyme design models to drug developers.
1. Core Product / Service
BioGeometry (北京百奥几何生物科技有限公司) was registered in Beijing on 2021-12-21; some databases give 2022 [1][2]. Its core model family is GeoFlow, which models atomic-level interactions between biomolecules. GeoFlow V2 (2024) was reported to match alphafold3 on antigen–antibody complex prediction [3]; the line has reached V3, and V4 is in development with the stated goal of moving from single-molecule design to cell-level "molecular system" design [4]. Earlier assets include TorchProtein, an open-source machine-learning library for protein work released in 2022, and PerturbDiff, a virtual-cell perturbation model [1]. Products are packaged as a generative antibody design platform (GeoBiologics) and an enzyme-engineering offering.
2. Target Users & Pain Points
- Antibody biotechs and pharma that want AI-generated binders, humanization and developability analysis without building an ML team.
- Antibody-discovery service companies that want an AI layer on top of their wet-lab capacity.
- Biomanufacturing customers needing enzyme optimization (stated use of the 2024 round) [3].
Business model. Three revenue lines are disclosed [1]: (1) model/platform licensing, where partners deploy the antibody discovery model and fine-tune it on their own project data; (2) fee-for-service work such as antibody humanization, developability assessment and affinity optimization; (3) co-development of pipeline projects with partners, which implies milestone or equity-style upside (terms undisclosed). Named licensees are Xiangyao Bio (祥耀生物, Dec 2023), which runs discovery, screening and activity assays itself, and Sanyou Bio (三优生物, Mar 2024), with which BioGeometry is building a combined dry–wet AIGC antibody platform [1]. A collaboration with Sino Biological on protein R&D has also been reported. Contract values and revenue are not disclosed.
Wet lab. Historically BioGeometry relied on partners' labs (Xiangyao, Sanyou) for validation. It now says it is building its own high-throughput wet-lab validation platform for macromolecules, and the June 2026 announcement describes its AI as "deeply integrated with automated laboratory platforms" [1]. Scale and location of that lab are not disclosed.
3. Competitive Landscape
| Player | Positioning | Vs. BioGeometry |
|---|---|---|
| moleculemind | MoleculeOS protein design platform | Closest peer; larger disclosed 2026 round |
| biomap | xTrimo foundation model, cell + protein | Broader scope and capital; IPO track |
| tianwu-tech | Protein-engineering services, enzyme-heavy | More service/industrial; BioGeometry more antibody |
| helixon-therapeutics | AI antibody design turned asset company | Shows the pipeline route peers can take |
| generate-biomedicines | US generative biologics with own clinic | Owns assets; BioGeometry mostly licenses |
4. Unique Observations
- Partnering with wet-lab-rich service companies was a capital-light way to get data. The Sanyou deal pairs BioGeometry's models with an antibody library and screening operation, so each partner project produces training data without BioGeometry funding the lab.
- The new round shifts it toward owning the loop. Building an in-house high-throughput lab and moving GeoFlow toward cells follows the wider pattern in China's AI-for-science market, where validation capacity, not model quality, is what customers pay for.
- Academic pedigree outside China. Tang Jian is a Mila (Montréal) professor; his LINE graph-embedding paper won a WWW 2026 Test of Time award [1]. Zhipu AI's participation in the 2024 round links the company to China's LLM ecosystem [3].
- Most performance claims are company-stated; no independent benchmark beyond published GeoFlow comparisons.
5. Financials / Funding
- Sep 2022, angel: tens of millions of USD, sole investor Gaorong Capital (高榕资本) [1].
- Jun 2024, Pre-A: led by Jiangmen Ventures (将门创投), with zhipu and 盛景嘉成; Gaorong followed on [3][5].
- Jun 2026, strategic round: "several hundred million RMB", co-led by the Shanghai biomedicine innovation-transformation fund (上实), 国科投资, Fortune Capital (达晨财智) and 星连资本; Gaorong and the 指数 AI industry fund followed [1][6]. One database labels it a ¥200M Series A (unverified).
- Valuation and revenue: not disclosed.
6. People & Relationships
- Founder: Tang Jian (唐建), associate professor at Mila / HEC Montréal; graph ML and AI-for-science researcher.
- Investors: Gaorong Capital, Jiangmen Ventures, zhipu, 达晨财智, 国科投资, 上实 biomedicine fund.
- Partners: Xiangyao Bio, Sanyou Bio, Sino Biological.
- Peers: moleculemind, biomap, tianwu-tech, xtalpi.
Sources: [1] pharnexcloud company profile; [2] China AI-drug company review 2026-10-14; [3] pedaily 2024-06; [4] 36kr 2026; [5] EqualOcean 2024-06; [6] Tencent News 2026-06-09.