Valhalla Technology has completed multiple rounds of financing totaling nearly 50 million US dollars, with Lighthouse Capital acting as the exclusive financial advisor.

Recently, Valhalla Tech, a paradigm-shifting leader in the AI4S track and a full-modality molecular world model enterprise, announced that it has successively completed multiple rounds of financing within three months, with a total amount of nearly 50 million US dollars. Investors include 5Y Capital, ZhenFund, CAS Star, Cheerhan Capital, Guofang Venture Capital, Lushi Investment, MiraclePlus, L2F Lighthouse Founders' Fund, and well-known industrial partners, with Lighthouse Capital acting as the exclusive financial advisor. The raised funds will be mainly used for expanding the talent team, iterating and upgrading the self-developed full-modality molecular world model, and continuously carrying out full-modality wet experiment verification, so as to further consolidate the company's full-stack technical foundation for Scientific AGI.

Trained under Professor David Baker, the 2024 Nobel Laureate in Chemistry, Zhang Haotian, founder of Valhalla Tech, comes from the Baker Lab, known as the "Huangpu Military Academy" in the global AI protein design field, where the Rosetta platform and the RFDiffusion generative model were born. Against the inertial trend of long-term focus on single-modality R&D in the AIDD industry, Zhang Haotian took the lead in breaking boundaries, proposing and constructing a brand-new paradigm of "full-modality molecular design". Its core breakthrough lies in starting from the underlying physical laws of molecules, describing all intermolecular interactions under a unified physical framework, thus overcoming the technical challenge of characterizing the commonalities and differences across cross-modality molecules.

Based on this original idea, the company is iteratively developing a new generation of multi-modal generative molecular world model - the AlloDesign system, which aims to achieve de novo design and structural optimization of any combination of the five major molecular modalities: proteins, DNA, RNA, small molecules, and ions, covering the full-chain molecular R&D needs. In addition, AlloDesign will systematically enhance the model's modeling and design capabilities for multi-modal and multi-dimensional interactions by pre-setting the physical assumptions and training process of many-body interactions, which is expected to break through the limitation of traditional single-modality models that cannot accurately simulate ternary and higher-order molecular interactions, helping to dig out a large number of "undruggable" targets that are difficult to develop with traditional technologies. At the same time, the company has built a supporting self-developed high-throughput experimental platform, aiming to open up the complete chain of molecular design, automated synthesis and affinity verification, so as to realize the rapid construction, experimental characterization and iterative optimization of candidate molecules. The experimental results will continuously feed back to the model and design strategies, promote the deep coupling of computational prediction and experimental measurement, and gradually form a data-driven closed-loop R&D system.

The AlloDesign system has taken the lead in achieving phased progress in the field of cyclic peptide design. Through technologies such as multi-modal hybrid training, random atomization and topological condition control, the model has achieved leading performance in benchmark tasks such as linear peptides and head-to-tail amide bond cyclic peptides, and further expanded to the de novo design of complex topological structures such as disulfide bond cyclization, isopeptide bond cyclization and bridged rings. Relevant progress shows that AlloDesign has initially possessed the unified modeling capability across different cyclization methods and chemical linkage types, laying a foundation for exploring the complex cyclic peptide chemical space that is difficult to cover by traditional methods.

In terms of commercialization, Valhalla Tech will take drug discovery as its first landing scenario in the short term, providing two types of services for global MNCs and Biotechs: customized target molecular R&D and industrial-grade model licensing, which will greatly shorten the pre-clinical R&D cycle of new drugs. In the medium and long term, the company will build a self-iterating Scientific AGI infrastructure, extend its technical capabilities to more scientific fields, and implement self-driving intelligent laboratories to create a virtual scientist system that can independently complete scientific deduction and experimental iteration, and build a truly underlying Scientific AGI platform for all fields.

Zhang Haotian, founder of Valhalla Tech, said: "We always believe that the overall technical route of AGI in Life Science must be from single-modality to multi-modality, and from single-task to multi-task. For molecular design, a full-modality model is not simply putting different molecular types into the same model, but hoping to form a unified technical system that can cross molecular modalities and connect computation and experiments starting from the underlying physical laws. After this round of financing, we will continue to invest in the construction of foundation models, high-throughput experimental platforms and top talents, and accelerate the transition of AlloDesign from cutting-edge research to real R&D scenarios."