Exclusive Interview with Dr. Song Yunlong: Over $7.8 Billion in Cumulative BD Deals, Marking the "Third" Venture of an AI Drug Discovery Pioneer

He has led his team in developing nearly 20 innovative small-molecule drug candidates, with multiple products advancing into Phase II and III clinical trials, generating over US$7.8 billion in business development (BD) deals to date:

After returning from Purdue University, he spent eight years conducting fundamental research at universities, during which he was selected for the Shanghai Pujiang Talent Program and received grants from the National Natural Science Foundation of China multiple times. In 2016, he moved to industry, holding positions at Medicilon, Fosun Pharma, GenSci and CSPC Pharmaceutical Group. He served successively as Director of Medicinal Chemistry, Head of Innovative Chemical Drug research and development at GenSci, General Manager of Shanghai Yishi Pharmaceutical, and Vice President of the Preclinical Business Division at CSPC Pharmaceutical Group.

In April 2026, he founded Yunji Zhiyao, shedding his accolades to start from scratch.

He is Dr. Song Yunlong.

A medicinal chemist well-versed in the end-to-end process of innovative drug development, from project initiation to clinical trials, and with two successful experiences building AIDD platforms from scratch, he now takes on a key question in his third startup: How can AIDD reconstruct the research and development paradigm for small-molecule innovative drugs?

In recent years, he has proposed that the rapid advancement of AI technology is driving a paradigm shift in new drug development (NDD)—from the traditional trial-and-error approach to a new model characterized by "AI modeling, automated experimentation and expert collaborative drive." In this transformation, AIDD-led guidance, dual-engine propulsion and rapid iteration are all indispensable; only by integrating them can the efficiency of NDD be significantly enhanced.

Yunji Zhiyao, the company he founded, is a biotech focused on the research and development of innovative small-molecule drugs. It is committed to empowering the entire new drug research and development chain with AI to create truly differentiated innovative therapeutics. Since its establishment, the company has rapidly integrated internal and external resources to build an end-to-end AI-driven drug discovery (AIDD) platform, covering the full drug discovery workflow including target assessment, dynamic protein modeling, molecule generation and ADMET prediction. Leveraging this platform, it has developed multiple high-value, differentiated small-molecule pipelines within just a few months.

Viewed over an extended timeline, while AI-driven drug discovery (AIDD) is important, it remains merely one component of the pharmaceutical industry. This methodological transformation has not overturned the research and development logic ultimately oriented toward clinical implementation. While platforms address efficiency, asset selection determines the future.

"Only when the direction is right do all efforts hold value. If the direction is wrong, all efforts will be in vain." This is how Song Yunlong summarizes a biotech's strategy for asset selection and future positioning.

Such is the pharmaceutical industry, and so too is the path of entrepreneurship. With favorable timing, geographic advantages and strong human support, he entered the field with full confidence and personal commitment.

The following content is edited by VCBeat.

VCBeat: You have served at multiple traditional pharmaceutical companies and led the discovery of several high-profile domestic small-molecule innovative drug projects. Why did you choose to launch a startup focused on "AI-driven discovery of high-value small-molecule innovative drugs" in 2026?

Song Yunlong: Over the past three years, China's innovative drug sector has made significant strides, with AI's role in research and development becoming increasingly tangible and several assets securing substantial overseas licensing deals. In 2026, the Government Work Report listed biopharmaceuticals as an "emerging pillar industry" for the first time, marking a favorable juncture for both the industry and its policy environment. As a leader in the Yangtze River Delta—the most dynamic hub for biopharmaceutical research and development in China—the team benefits from a highly advantageous research and development ecosystem. Core team members have collaborated across various platforms for many years, fostering deep mutual trust.

Therefore, 2026 proved to be a year with all favorable conditions—timing, location, and human harmony—in place. In April, Yunji Zhiyao was registered and established in Shanghai.

As for strategic direction, small molecules remain the dominant modality in innovative drug development. However, traditional medicinal chemistry is characterized by high trial-and-error costs and long development cycles, while achieving true Me-better or Best-in-class status demands higher technical barriers. Leveraging my prior experience in medicinal chemistry, combined with two instances of building AI-driven drug discovery (AIDD) platforms from the ground up, I am well-positioned to utilize end-to-end AI capabilities to accelerate and enhance the precision of discovering high-value, differentiated small molecules. Our choice of this direction reflects our commitment to undertaking difficult yet correct endeavors.

VCBeat: How do you view your current entrepreneurial venture?

Song Yunlong: I experienced two instances of "simulated entrepreneurship" at my previous two companies—both starting from scratch or during trough periods, where I built platforms while executing projects. Apart from the absence of fundraising pressure, I faced nearly all the typical challenges of entrepreneurship. This time, it is a true startup journey from the ground up, which I prefer to call my third entrepreneurial venture in life. Different platforms bring different challenges, as well as distinct experiences and gains. I am eagerly looking forward to this new journey.

VCBeat: As a pharmaceutical development team, what is the core problem that Yunji Zhiyao aims to solve?

Song Yunlong: Great truths are simple. The problem we aim to solve is straightforward: to deliver truly effective medicines for the general public—and we already possess the fundamental conditions to achieve this goal.

Yunji's core team consists of old friends who have worked together for many years. In just the past four to five years, we have developed 12 preclinical candidate drugs and secured 14 clinical trial approvals in China and the United States, with multiple projects advancing into Phase I and Phase II clinical trials, demonstrating promising therapeutic potential.

Next, we aim to integrate our years of accumulated research and development experience with the latest advancements in AI-driven drug discovery (AIDD). Leveraging a full-chain AI platform and relying on the new drug research and development ecosystem in the Yangtze River Delta, we strive to develop safer and more effective innovative drugs to treat diseases and save lives. This goal may not sound particularly grandiose, but we believe it is the ultimate "touchstone" for testing all those involved in drug development.

VCBeat: How does the logic of AIDD-native companies differ from the linear, engineering-driven approach of traditional pharmaceutical companies?

Song Yunlong: The development of innovative drugs is a high-risk, long-cycle systematic engineering endeavor that is science-driven, clinically oriented, and subject to global competition. The research and development organizations of traditional pharmaceutical companies often bear the imprint of the generic drug era, characterized by stability and cost-control proficiency; however, an exclusive pursuit of risk minimization and cost reduction often comes at the expense of quality and efficiency.

"AI-Native" is not merely a technical label; more importantly, it represents a reconstruction of research and development paradigms and organizational structures. The value of AI lies not in isolated tools, but in enabling data feedback from every stage of the "design–synthesis–test–analysis" cycle, thereby achieving closed-loop data flow across the entire pipeline to continuously refine models through iterative improvement.

What AI-native companies need to do is return organizations to the fundamental principles of R&D: establishing a full-link data closed loop with feedback and iteration, and designing organizational structures accordingly. This approach is based not only on our understanding of R&D dynamics but also on our years of firsthand experience in traditional enterprises. We aim to achieve faster response times, more decisive decision-making, and higher efficiency through the AI-native transformation of organizations. With full-link AI empowerment and continuous model optimization, costs will also be optimized in the long run.

VCBeat: Built on an AI-native organizational structure, what are the key features and differentiators of Yunji Zhiyao's self-developed end-to-end AI platform?

Song Yunlong: Our AI platform has three features:

First, we were nurtured in real-world research and development scenarios. Our past experiences in platform development and pipeline research and development, regardless of success or failure, have been systematically reviewed and consolidated into data and methodologies. Our approach is the reverse of that taken by most AI companies: we first rooted ourselves in authentic research and development environments before building tools to address specific challenges. Consequently, the objective function of our model is inherently aligned with "how to develop a drug into a marketable therapy," rather than merely achieving high scores on benchmark tests.

Second, a unified data foundation is shared across the entire pipeline. The stages of design, synthesis, testing, and analysis (DMTA) are mutually intelligible, with experimental conclusions from each cycle fed back to train the models, thereby systematically compressing the DMTA turnaround time.

Third, we deeply integrate human-AI division of labor. AI has proven useful in hit generation, but the greater value lies in progressing from hits to high-quality PCCs, as wet-lab throughput lags far behind AI's generation speed.

Thus, our division of labor is: humans set the direction, AI generates molecules, and humans make decisions—scientists define the design direction based on biological insights, while generative AI produces candidate molecules in bulk; these are then evaluated and filtered by physics-based CADD and data-driven AIDD systems, upon which scientists make the final decision to select target molecules for wet-lab validation.

This leads to two key differentiators:

Shift Failure Upstream—calculate ADMET, toxicity and synthesizability early in project initiation to turn costly failures into affordable ones;

Closed-Loop Acceleration—the cycle from hypothesis to the next round of experiments is shortened by an order of magnitude compared with traditional organizations.

The true quality of a platform ultimately depends on the caliber of its pipeline.

VCBeat: What specifically should be handed over to AI? What should be left to humans?

Song Yunlong: AI is leveraged to the greatest extent possible across all stages, yet strategic direction and critical decision-making remain in human hands. Humans must also systematically evaluate the capabilities of AI at each stage and determine the appropriate weight to assign to its outputs. Current AI algorithms are no longer the primary bottleneck. The success of large language models hinges on the availability of vast amounts of high-quality data; however, new drug development is precisely constrained by a scarcity of such data, compounded by the difficulty of sharing proprietary internal data among pharmaceutical companies. This determines that current AIDD models remain "correct to a limited extent."

Acknowledging this reality allows for more clear-headed and effective utilization of AI. While AI can suggest probabilities, it cannot yet provide certainty and is prone to hallucinations. Therefore, while leveraging AI to its fullest extent, key decisions should remain in human hands, with data continuously validated through wet-lab experiments to drive rapid model iteration. This approach not only accelerates the discovery of high-quality molecules but also facilitates the accumulation of high-quality data, thereby enhancing the accuracy of AI models over time.

VCBeat: Previously, you shared that it took only five weeks to complete the molecular design, synthesis, and initial screening of a high-value autoimmune target. Could you please elaborate on this case?

Song Yunlong: This is a popular target in the autoimmune field, which we have been monitoring for some time. During the project discussion on June 9, our team members brought up this target. After deliberation, we concluded that it holds significant clinical and business development (BD) value. Meanwhile, its mechanism of action remains unclear; although existing competitor molecules have completed preliminary clinical proof-of-concept, there is still substantial room for improvement.

That afternoon, the AI project initiation agent completed system intelligence research and assessment conclusions within 10 minutes; patent colleagues leveraged an AI patent analysis platform to rapidly analyze competitors' published patents; medicinal chemistry colleagues utilized an AI molecular design platform to swiftly conduct SAR (Structure-Activity Relationship) analysis and mechanism of action analysis on existing patented molecules, thereby forming preliminary molecular design strategies.

Subsequently, we held another meeting to review the findings from project initiation research, patent analysis, SAR summaries, and molecular design. Before the end of the workday, the first batch of designed molecules had been sent to our collaborative CRO, who immediately initiated route design and reagent procurement.

The target molecules for this project were not difficult to synthesize. Taking into account the actual procurement time, the first batch of molecules was synthesized within two weeks, and the activity tests yielded excellent results. Patent layout was completed on the same day, and ADME evaluation was arranged for the molecules with superior activity.

Following the second and third rounds of optimization, by July 7—spanning approximately 20 working days—we not only identified a molecule with an enzyme inhibitory activity of approximately 1 nM, but also demonstrated that it was comprehensively superior to current clinical competitor molecules in terms of potency, selectivity, and CYP inhibition. On the same day, we held a meeting with business development contacts from two potential multinational corporation (MNC) buyers to preliminarily discuss project details and further confirm their requirements.

This molecule has demonstrated strong PCC potential, laying a solid foundation for patent positioning, SAR breakthroughs, and further design, with in-depth biological evaluation currently underway. Building on this progress, we have optimized a new molecule with nearly tenfold improved potency. We aim to develop the globally best-in-class molecule for this target and advance it into clinical development.

VCBeat: Are there any cases involving failure or phased failures? For instance, situations where in silico experiments proceeded smoothly, but wet lab experiments proved contradictory?

Song Yunlong: Of course. Occasionally being proven wrong is not necessarily a "bad thing."

On one occasion, our AI designed a theoretically "perfect" novel scaffold molecule. The model analysis strongly supported it, but I had some reservations at the time—this scaffold had never been explored before, and its structure-activity relationship (SAR) remained unvalidated. Would it actually work? After discussion, the team concluded: since the model analysis was supportive, we should give it a try.

Our CRO partners devoted considerable effort to successfully synthesize the target molecule, only to find it inactive in biological assays. Through in-depth AI analysis, the team identified potential underlying causes, significantly enhancing their understanding of the target and its binding mode. This insight enabled the rapid design of superior molecules for the next round.

Scientific exploration is never smooth sailing. In the short term, it may seem like setbacks, but once the issues are thoroughly understood, it will yield better results. Currently, AIDD has not yet reached a stage where it can directly provide definitive outcomes. This "limited correctness" remains meaningful, representing the current state of AI-driven drug discovery (AIDD).

VCBeat: Can the upfront payment and royalty sharing for AIDD be negotiated to more favorable terms, and does this depend on whether the counterparty recognizes the value of AIDD?

Song Yunlong: Multinational pharmaceutical companies now widely recognize AI-driven drug discovery (AIDD) and are actively making strategic investments. However, if the transaction target is a drug pipeline, whether favorable pricing can be negotiated remains uncertain. Ultimately, it depends on clinical needs, target value, molecular differentiation, and the competitive landscape, with preliminary clinical data being the most critical factor.

If the subject of the transaction is research and development collaboration, the situation will be different. A fully validated AIDD platform can instill greater trust in partners: it does not rely on any single scientist, but instead leverages the platform to achieve batch production of high-quality molecules.

Yunji Zhiyao has been established for only a few months, yet multinational corporations (MNCs) and overseas funds have already approached the company to explore potential collaborations. Our primary strategy is to pursue business development (BD) transactions for international rights once our in-house research and development reaches a certain stage, while also considering research and development partnerships with strategic allies.

VCBeat: What do MNCs focus on in early-stage pipelines?

Song Yunlong: MNCs' current focus in China on early-stage assets is primarily on Me-better or Best-in-class candidates with differentiated advantages: First, the target has preliminarily completed clinical proof-of-concept overseas; second, the China project demonstrates significant differentiated advantages in the preclinical stage; third, it involves cornerstone therapies in key therapeutic areas or candidates capable of generating synergistic value with our core pipeline.

Meanwhile, multinational corporations (MNCs) are willing to pay a premium for differentiated advantages—not merely a 3- to 5-fold increase in potency at the molecular level, but rather a qualitative improvement in animal efficacy, a substantial enhancement in safety, or a breakthrough in dosing convenience. The recent deal for PepStar Therapeutics' oral IL-23R cyclic peptide, with an upfront payment as high as US$1.275 billion, serves as a prime example.

VCBeat: Clinical data, especially late-stage data, remains the most critical criterion. Does this mean that AIDD still adheres to the traditional logic of pharmaceutical business development (BD)?

Song Yunlong: Indeed. AIDD does not change the standards, but rather the means by which they are achieved.

Clinical data, particularly late-stage data, remains the most critical currency in business development (BD) pricing, and AI-driven drug discovery (AIDD) has not changed this. The pipeline valuation frameworks of multinational corporations (MNCs) assess targets, data, competitive landscape, and development risks; they do not assign a premium simply because a molecule was AI-generated.

Research and development collaborations based on AIDD platforms typically prioritize upfront payments and service and cooperation returns. Whether these platforms can consistently generate high-quality molecules, and whether such molecules can successfully navigate subsequent development stages, remains to be validated. Given the high uncertainty surrounding the value generated, valuing companies solely based on their platform capabilities poses significant challenges. Therefore, to secure larger profit-sharing arrangements, the focus must ultimately return to pipeline value—where international markets have more mature valuation frameworks for new drug pipelines.

VCBeat: This is also a major topic of discussion in AI-driven drug discovery (AIDD): whether to conduct clinical trials independently and advance candidates to later stages, or to pursue international expansion at an early stage.

Song Yunlong: Currently, the vast majority of Chinese biotech companies lack the capability to commercialize their products overseas. Business development (BD) is a viable and effective pathway to monetize the international rights of assets that these companies are not equipped to commercialize on their own.

Going global early does not mean easily abandoning the Chinese market; the Chinese rights to core assets should, as much as possible, remain in the hands of Chinese entities. If several competing products already exist in the Chinese market and future competition is expected to become intense, directly transferring global rights is also a sound strategy. For instance, our decision to consider an early transfer of the global rights to our internally developed small-molecule Lp(a) project was driven by several factors: first, there were already multiple small-molecule candidates targeting the same mechanism in China, with at least two competitors ahead in development; second, our group had also laid out its own siRNA portfolio targeting the same mechanism; and third, the Phase III clinical trials for this asset would require substantial capital investment and time.

Should companies go global early, or conduct clinical trials domestically and expand overseas in the mid-to-late stages? Not only potential buyers but also the competitive landscape should be considered. For products with high barriers to entry and a leading competitive position, it is advisable to defer commercialization as late as possible, ideally until Phase II efficacy data are available, to maximize returns.

VCBeat: AI projects now need to demonstrate value realization and subsequent implementation. How do you view the value realization of AIDD?

Song Yunlong: In previous years, biotech companies primarily relied on advancing products to Phase II or III before going public to realize value. Over the past three to five years, promising Phase I and II data have enabled value realization through business development (BD) deals, with milestones occurring increasingly earlier. In recent years, several leading Chinese AI-driven drug discovery (AIDD) companies have signed numerous high-value transactions with multinational corporations (MNCs), accelerating the timeline for realizing AI's value in pharmaceutical development.

In 2024, when we negotiated with AstraZeneca over our Lp(a) inhibitor, the molecule had not yet been nominated as a preclinical candidate (PCC), yet we secured a US$100 million upfront payment. The key drivers were unmet clinical needs, the target's value and risk profile, and preclinical data demonstrating the molecule's differentiated advantages.

At Yunji Zhiyao, we uphold high research and development standards for our projects. Our primary focus is to deliver maximum value at the right time and in the most optimal manner, though this ultimately depends on each project's circumstances. Although some projects start early, competitors may quickly develop me-too products, necessitating the earliest possible value realization. In contrast, for products with high technical barriers, broad clinical demand and the potential to disrupt existing treatment outcomes, it is advisable to advance toward international clinical trials—provided the competitive landscape remains unchanged—to seek greater value realization in the global market.

For example, the GenSci074 project I previously worked on at GenSci has entered international multicenter Phase III clinical trials. With significant improvements in both clinical efficacy and safety, its business development (BD) value is substantially enhanced. I look forward to seeing this product successfully launched in overseas markets soon.

VCBeat: Overall, how do you anticipate the commercialization path for Yunji Zhiyao?

Song Yunlong: Biotech companies must first ensure their survival, as only by surviving can they secure a future. Their greatest advantages lie in efficiency and flexibility, and they remain open to all commercialization models, including conventional platform collaborations and business development (BD) license-out deals.

Yunji Zhiyao, both now and in the future, will always welcome any collaborative model characterized by mutual benefit and complementary advantages: whether collaborating with multinational corporations (MNCs) or domestic pharmaceutical companies with strong clinical development or commercialization capabilities in specific disease areas, jointly initiating projects, or even directly undertaking targets of interest to partners to leverage each other's strengths. The collaboration model does not involve charging a hefty one-time research and development fee; instead, it fully accounts for the partner's risk by charging moderate early-stage research and development costs and milestone payments, supplemented by reasonable downstream sales royalties, thereby fostering a mid-to-long-term partnership that shares both risks and rewards.

At the appropriate juncture, after accumulating greater resources, we aim to hold domestic rights to several core assets and transition into a biopharma company at the right time, creating value for shareholders and making greater contributions to society.

VCBeat: How has your extensive prior BD experience shaped Yunji Zhiyao's expansion strategy?

Song Yunlong: Global expansion is the top priority of our research and development efforts. Past transactions have made one thing clear: valuable pipelines do not need to wait until fully mature before going global; early-stage assets can also secure significant cross-border collaborations. Therefore, over the next 1–3 years, Yunji Zhiyao will make global expansion the core pathway for value realization. High-quality pipelines initiated in-house will also be advanced for overseas licensing at appropriate times, allowing us to both accumulate resources and support our broader future ambitions.

Over the years, we have executed multiple business development (BD) transactions, with related projects demonstrating significant differentiated advantages at the clinical stage. This has not only validated the strength of our team and research and development platform to the market but also built a solid foundation of trust and reputation. We believe this track record and established credibility will enhance partner confidence and accelerate the global expansion of Yunji Zhiyao's product pipeline.

VCBeat: Under the guidance of its overseas expansion strategy, how is Yunji Zhiyao structuring its pipeline and indications?

Song Yunlong: In terms of strategic layout, we must first consider: Who are the target customers? Who are the end customers? By clarifying these questions, we can work backward from both ends—determining what multinational corporations (MNCs) need and what patients truly need. The principle for project initiation is to possess high technical barriers and differentiated advantages, address unmet clinical needs, overcome the significant shortcomings of existing standard treatments, and have the opportunity to rank among the top three internationally in a specific niche sector.

Commercial value is invariably underpinned by clinical needs; however, the presence of a clinical need does not necessarily translate into an immediately viable business opportunity. If existing therapies already demonstrate high efficacy, a first-in-class drug may represent merely a scientific innovation rather than one delivering tangible patient benefits or becoming a blockbuster product.

VCBeat: What progress and directions can be disclosed regarding the pipeline?

Song Yunlong: There remains a significant unmet need in chronic diseases, particularly autoimmune and inflammatory conditions: either existing treatments offer suboptimal efficacy, as seen in inflammatory bowel disease, or there is a lack of available therapies for certain diseases. Consequently, this area has become a focal point for innovative drug development among multinational pharmaceutical companies.

We have strategically laid out our initial portfolio of several core targets in this field, all of which have just completed clinical proof-of-concept, address large indications, and offer significant room for improvement over existing competitors. The development of these pipelines is progressing smoothly, having overcome key technical bottlenecks, and we expect to complete the molecular discovery phase in the near future.

VCBeat: Why the natural shift toward the autoimmune disease sector?

Song Yunlong: While there are already numerous drugs in the autoimmune field, the mainstream options are predominantly large-molecule biologics, most of which target inflammatory cytokines and provide symptomatic relief. The most significant challenge is the need for long-term, continuous administration; upon discontinuation, the disease often rebounds rapidly, making it difficult to sustain efficacy. Our goal is twofold: first, to offer patients a more convenient oral option that effectively controls the disease; and second, to move beyond mere symptom management by developing therapies that maintain efficacy even after discontinuation, thereby addressing the root cause.

VCBeat: Does autoimmunity want to become the label of Yunji Zhiyao?

Song Yunlong: The first wave of projects indeed all fell under the broad category of autoimmune or inflammatory diseases. However, this was not the strategic direction chosen at the outset of the venture; rather, it represented a fortuitous alignment between the overarching therapeutic area and specific market opportunities.

However, Yunji Zhiyao does not wish to be defined solely as an autoimmune company. Over the years, we have deeply cultivated nearly all therapeutic areas. We prefer to position ourselves as a company dedicated to addressing unmet and urgent clinical needs in chronic diseases by developing novel therapeutics with differentiated advantages, or even next-generation innovative drugs.

VCBeat: How to understand "second-generation chronic disease products"?

Song Yunlong: The so-called second-generation products refer to those targeting a therapeutic area where treatment value has been preliminarily proven both domestically and internationally, but the existing molecules are not optimal, leaving significant room for improvement. Our goal is to maximize the therapeutic potential of these targets and develop truly superior second-generation products, rather than creating me-too versions of existing drugs. Many of our past projects or deals, such as ROCK2 inhibitors, NK3R antagonists and Lp(a) inhibitors, have demonstrated the potential to become second-generation products, proving that our team possesses this capability.

VCBeat: Can it be understood that, empowered by AI, engineering innovation can be executed better and more effectively?

Song Yunlong: When it comes to Chinese innovation, people tend to think of engineering capabilities, cost-effectiveness and high efficiency. Indeed, certain areas within the innovative drug sector require strong engineering capabilities, such as ADCs (antibody-drug conjugates) and siRNA (small interfering RNA). While developing me-too drugs may not be particularly challenging, achieving genuine differentiated advantages is far from easy.

In contrast, small molecules are still far from being amenable to engineered innovation. It is often said that small-molecule design is an art; although the advent of AI-driven drug discovery (AIDD) has improved efficiency and reduced costs, significant technical barriers remain. This is a key reason we have chosen this direction—to focus on what we are passionate about, excel at, and where high technical barriers exist.

VCBeat: Why haven't you completed your first round of financing yet?

Song Yunlong: We are deeply grateful to our friends at various institutions for their recognition of me and the Yunji Zhiyaoteam. Through our exchanges with these partners, I have gained deeper insights into the entrepreneurial journey, drawing from years of hands-on experience on the front lines of research and development. While we highly value and are always willing to listen to the suggestions and feedback from our institutional partners, we still wish to maintain a high degree of autonomy over our research and development direction.

Therefore, although I always had some options at hand, I decided to run the first leg.

During the fundraising process, several overseas funds approached us sequentially, seeking to collaborate on development or jointly establish a Newco. We once considered whether we could postpone urgent fundraising if we could secure capital from our partners.

Over time, we have increasingly recognized that entrepreneurship is a long-term endeavor requiring resources and support across various dimensions, not merely capital. With multiple projects now advanced to significant stages and preliminary data demonstrating clear advantages, it has become an opportune moment to bring on board partners who share our vision. We are rapidly advancing our Series A financing round, and everything is progressing smoothly.

VCBeat: What kind of AI story does Yunji Zhiyao aim to tell?

Song Yunlong: Yunji Zhiyao is a biotech focused on the development of differentiated small-molecule innovative drugs. Rather than developing algorithms from scratch or chasing model parameters, we fully leverage leading AI tools in the industry, customizing and integrating them into real-world scenarios. By embedding drug discovery logic and decision-making methodologies into these systems, we establish a closed loop connecting data, models and research and development practices, enabling continuous iteration and optimization throughout the actual process of new drug development.

Our most valuable asset is a drug development methodology, validated through nearly 20 PCC molecules, over a dozen regulatory approvals, and multiple Phase II/III clinical trials of Me-better/Best-in-class candidates. This methodology covers the entire process from project initiation to value realization, addressing key questions such as target selection, molecular design, sources of differentiation, risk control, and value maximization. Combined with practical reviews of multiple BD transactions, this methodology forms the “benchmark” by which we evaluate and screen AI-generated molecules.

Having engaged with computer-aided drug design (CADD) and medicinal chemistry for nearly three decades, I have witnessed the complete lifecycle of CADD—from its nascent stages and subsequent hype cycle to its trough of disillusionment and eventual resurgence. Progressing from molecular mechanics and molecular dynamics to quantum chemistry, I have developed an understanding of intermolecular interactions at the electronic level. Furthermore, through the research and development of nearly 20 preclinical candidate (PCC) molecules, I have gradually achieved a deep integration of theoretical computation with practical medicinal chemistry.

The advent of the AI era has revitalized computer-aided drug design (CADD), offering a glimpse of breakthrough potential and motivating my active engagement in AI-driven drug discovery (AIDD). Years of accumulated knowledge and experience enable us to evaluate AI-designed molecules with both the depth of theoretical computation and the breadth of medicinal chemistry practice.

We are a team of pharmaceutical industry veterans with years of deep expertise in China's pharmaceutical sector. By leveraging end-to-end AI empowerment, we aim to efficiently develop truly high-quality medicines, advance clinical trials, and realize international value through business development (BD) transactions, ultimately benefiting patients worldwide. This is the AI story of Yunji Zhiyao.