Artificial intelligence (AI) is rapidly transforming the landscape of drug discovery, as pharmaceutical and biotech companies increasingly seek innovative solutions to compress lengthy research and development (R&D) timelines and enhance success rates amidst escalating costs. This burgeoning field has attracted over 200 startups, all vying to embed AI directly into research workflows, drawing significant attention and investment.

Converge Bio, a Boston- and Tel Aviv-based startup, stands at the forefront of this shift, having recently secured a substantial $25 million in an oversubscribed Series A funding round. This new capital underscores the intensifying competition within the AI-driven drug discovery sector. The round was led by Bessemer Venture Partners, with additional participation from TLV Partners, Saras Capital, and Vintage Investment Partners, alongside backing from undisclosed executives at industry giants like Meta, OpenAI, and Wiz.

Converge Bio specializes in accelerating drug development for pharma and biotech clients by leveraging generative AI, which is meticulously trained on vast datasets of molecular information, including DNA, RNA, and protein sequences. As articulated by Converge Bio CEO and co-founder Dov Gertz, their platform is designed to support various defined stages of the drug-development lifecycle, from target identification and discovery through manufacturing and clinical trials. This holistic approach aims to bring new drugs to market with unprecedented speed.

The company has already deployed several customer-facing AI systems tailored to specific needs. These include a sophisticated system for antibody design, another for optimizing protein yield, and a third for biomarker and target discovery. Gertz elaborated on the antibody design system, explaining it's not merely a singular model but an integrated suite of three components: a generative model for creating novel antibodies, predictive models for filtering these antibodies based on molecular properties, and a physics-based docking system that simulates three-dimensional interactions between antibodies and their targets. Gertz emphasized that the true value lies in these comprehensive, ready-to-use systems that seamlessly integrate into existing workflows, rather than individual models that customers would have to piece together themselves.

Converge Bio's rapid ascent is evident in its operational growth since its $5.5 million seed round in 2024. In just two years, the startup has completed over 40 programs with more than a dozen pharmaceutical and biotech customers across the U.S., Canada, Europe, and Israel, with plans for expansion into Asia. Its team has expanded significantly, growing from nine employees in November 2024 to 34. Public case studies illustrate the platform's impact, demonstrating achievements such as boosting protein yield by 4 to 4.5 times in a single computational iteration and generating antibodies with exceptionally high binding affinity, reaching the single-nanomolar range.

The broader industry context reflects a strong wave of interest in AI-driven drug discovery. Notable developments include Eli Lilly's collaboration with Nvidia to build a powerful supercomputer for drug discovery, and the developers of Google DeepMind's AlphaFold receiving a Nobel Prize in Chemistry for their AI system that predicts protein structures. Gertz highlighted that the industry is experiencing the largest financial opportunity in life sciences history, shifting away from traditional “trial-and-error” methods towards data-driven molecular design. He noted that initial skepticism about AI in drug discovery has rapidly dissipated due to successful case studies from both companies like Converge and academic institutions.

While large language models (LLMs) are garnering attention for their potential in analyzing biological sequences and suggesting new molecules, challenges such as hallucinations and validation costs remain significant. Gertz pointed out that unlike text, where hallucinations are often easy to spot, validating a novel molecular compound can take weeks, incurring much higher costs. To address this, Converge Bio employs a strategy of pairing generative models with predictive ones, effectively filtering new molecules to reduce risk and enhance outcomes for its partners. Gertz also concurred with experts like Yann LeCun regarding the limitations of text-based LLMs for core scientific understanding, asserting that models must be trained on fundamental biological data like DNA, RNA, proteins, and small molecules. Text-based LLMs are used by Converge only as supportive tools, such as aiding customers in navigating literature on generated molecules, and are not central to their core technology. The company maintains a flexible approach, utilizing various architectures including LLMs, diffusion models, traditional machine learning, and statistical methods as appropriate.

Converge Bio's ultimate vision is to establish itself as the generative AI lab for every life-science organization. Gertz envisions a future where traditional "wet labs" will always exist but will be synergistically paired with "generative labs" that computationally create hypotheses and molecules. This ambition positions Converge Bio to become the foundational generative lab for the entire life sciences industry.