As pharmaceutical and biotechnology companies seek methods to shorten R&D cycles and improve success rates to combat ever-increasing research costs, artificial intelligence is rapidly making inroads into the drug discovery field. Over 200 startups are now racing to deeply integrate AI technology into their R&D processes, attracting growing attention from investors. Amid intensifying competition in the AI-driven drug discovery arena, Converge Bio has emerged as the latest player in this wave, successfully securing a new round of funding.
This startup, headquartered in Boston, USA, and Tel Aviv, Israel, leverages generative AI technology trained on molecular data to help pharmaceutical and biotech companies accelerate their drug development processes. The company has completed an oversubscribed $25 million Series A funding round, led by Bessemer Venture Partners, with participation from TLV Capital and Classic Investment Partners. Additional investment was also provided by anonymous executives from Meta, OpenAI, and Wiz.
In practical application, Converge Bio trains generative models based on DNA, RNA, and protein sequences, subsequently integrating these models into the R&D workflows of pharmaceutical and biotech companies to speed up drug development.
Converge Bio's CEO and Co-founder, Dov Gertz, stated in an exclusive interview: "The drug discovery cycle has distinct phases – from target identification and discovery to manufacturing, clinical trials, and more. Our technology can support relevant experiments for each of these stages. Our platform is continuously expanding to cover these R&D phases, helping new drugs reach the market faster."
To date, Converge Bio has launched customer-facing application systems, specifically comprising three independent AI systems designed for antibody design, protein yield optimization, and biomarker and drug target discovery, respectively.
Gertz elaborated further: "Take our antibody design system as an example; it's not a single model but consists of three interconnected components. First, a generative model creates novel antibodies. Second, a predictive model filters them based on molecular properties. Finally, a physics-based molecular docking system simulates the three-dimensional interaction between an antibody and its target." He emphasized that the system's core value lies in its integrated, synergistic operation rather than relying on any single model. "Clients don't need to integrate various models themselves; they get a ready-to-use system that plugs directly into their own R&D workflows."
This funding round comes approximately a year and a half after the company completed a $5.5 million seed round in 2024.
Gertz revealed that this two-year-old startup has experienced rapid expansion. Converge Bio has now established partnerships with 40 pharmaceutical and biotechnology companies, with around 40 R&D projects running concurrently on its platform. Its business spans the US, Canada, Europe, and Israel, and the company is actively working to expand into the Asian market.
The company's team size has also grown rapidly, expanding from 9 people in November 2024 to 34 today. During this period, Converge Bio has also published several public case studies. Gertz pointed out that one case demonstrated how the company helped a partner increase protein yield by 3 to 3.5 times in a single computational iteration; in another case, antibodies generated by its platform exhibited extremely high binding affinity, reaching the nanomolar level.
The field of AI-driven drug discovery is experiencing an investment boom. Last year, Eli Lilly partnered with NVIDIA to create what is claimed to be the most computationally powerful drug discovery supercomputer in the pharmaceutical industry. In October 2024, the research team behind Google DeepMind's AlphaFold project, responsible for this AI system that predicts protein structures, was awarded the Nobel Prize in Chemistry.
When asked about the impact of this industry momentum on Converge Bio's growth, Gertz stated that the life sciences sector is facing the largest commercial opportunity in its history, as the industry transitions from traditional "trial-and-error" R&D models to data-driven molecular design.
"Witnessing this boom is something we feel very tangibly, evident just from looking at our company's inbox," Gertz shared. "A year and a half ago when the company started, there was still a lot of skepticism in the industry." He added that, thanks to successful case studies published by Converge Bio and other companies, as well as from academia, this skepticism has rapidly dissipated.
While large language models have garnered significant attention in drug discovery for their ability to analyze biological sequences and design novel molecules, the technology still faces challenges such as hallucination effects and insufficient accuracy. "In the text domain, hallucinations are often easy to spot; but in molecular R&D, validating a novel compound can take weeks, making the cost of trial and error much higher," Gertz explained. To address this, Converge Bio employs a strategy that combines generative models with predictive models, filtering novel molecules to reduce R&D risks and enhance partners' efficiency. "This filtering mechanism, while not perfect, significantly reduces risk and delivers better outcomes for our clients."
Gertz was also asked about his views on experts like Yann LeCun, who remain skeptical about the application of large language models. "I am a big admirer of Yann LeCun and completely agree with his perspective," Gertz clarified. "We do not rely on text-based models for core scientific understanding. To truly grasp biological mechanisms, models must be trained on DNA, RNA, protein, and small molecule data."
He further clarified that text-based large language models serve only as auxiliary tools within the company's technology stack, for instance, by helping clients retrieve literature related to generated molecules. "Such models are not our core technology. We are not wedded to a single technical architecture; instead, we flexibly utilize large language models, diffusion models, traditional machine learning, and statistical methods as needed by the task at hand."
Outlining the company's vision, Gertz stated: "We aspire to be the generative AI research lab for all life sciences organizations. Physical wet labs will continue to exist, but they will operate in synergy with generative labs that computationally generate hypotheses and molecules. Our ambition is to become the generative lab for the entire industry."
Comments