At a Glance
- Converge Bio closed an oversubscribed $25 million Series A led by Bessemer Venture Partners
- The Boston-Tel Aviv startup trains generative AI on DNA, RNA, and protein sequences to accelerate drug pipelines
- The two-year-old company already runs 40 programs with 40 pharma and biotech partners
- Why it matters: Faster, data-driven molecule design could shorten costly R&D timelines and boost success rates for new medicines
AI is racing into drug discovery as more than 200 startups compete to compress R&D timelines and raise approval odds. Converge Bio, which builds generative models for molecular data, has secured fresh capital to expand its platform.
Oversubscribed Series A
Converge Bio raised $25 million in an oversubscribed Series A. Bessemer Venture Partners led the round, joined by TLV Partners and Vintage Investment Partners. Executives from Meta, OpenAI, and Wiz also invested.
The financing arrives roughly 18 months after the company’s $5.5 million seed raise in 2024.
How the Platform Works
The startup trains generative models on DNA, RNA, and protein sequences, then embeds them into pharma and biotech workflows. Customers receive ready-to-use systems rather than standalone models.
Chief executive and co-founder Dov Gertz broke down the antibody design system:
- A generative model creates novel antibodies
- Predictive models filter candidates by molecular properties
- A physics-based docking simulator tests 3-D interactions
Gertz emphasized that value lies in the integrated system, not any single component.
Customer Systems Live Today
Converge has released three AI systems:
- Antibody design
- Protein yield optimization
- Biomarker and target discovery
The company has signed 40 partnerships and is running about 40 programs across the U.S., Canada, Europe, and Israel. Expansion into Asia is under way.
Rapid Scaling
Headcount has jumped from nine employees in November 2024 to 34. Public case studies show early wins:
| Metric | Result |
|---|---|
| Protein yield | 4-4.5× boost in one computational iteration |
| Antibody binding | Single-nanomolar affinity achieved |
Industry Momentum
Major players are doubling down on AI-driven discovery:
- Eli Lilly partnered with Nvidia to build what they call the pharma industry’s most powerful supercomputer for drug research
- Google DeepMind’s AlphaFold team won the 2024 Nobel Prize in Chemistry for protein-structure prediction
Gertz told News Of Philadelphia the field is witnessing “the largest financial opportunity in the history of life sciences,” shifting from trial-and-error to data-driven molecular design. Skepticism that greeted the company 18 months ago has “vanished remarkably quickly,” he added.
Handling Hallucinations
Large language models can analyze biological sequences, yet hallucinations carry higher costs in molecules than in text. Validating a novel compound can take weeks.
Converge pairs generative models with predictive filters to reduce risk. “This filtration isn’t perfect, but it significantly reduces risk and delivers better outcomes for our customers,” Gertz said.
The company limits text-based LLMs to support tasks such as literature navigation. Core scientific understanding comes from models trained on DNA, RNA, proteins, and small molecules.
Tech Stack Flexibility
Converge does not rely on a single architecture. The platform deploys:
- LLMs
- Diffusion models
- Traditional machine learning
- Statistical methods
Gertz said the firm’s vision is to become “the generative AI lab” for every life-science organization, complementing rather than replacing wet labs.
Key Takeaways
- Converge Bio’s $25 million Series A signals continued investor appetite for AI drug discovery
- The company’s integrated systems approach addresses accuracy and workflow adoption concerns
- With 40 active programs and rapid team growth, Converge is scaling ahead of many peers
- Generative molecular design is moving from promise to measurable R&D impact
