Converge Bio raises $25M, backed by Bessemer and executives from Meta, OpenAI and Wiz

Artificial intelligence is moving rapidly into drug discovery, as pharmaceutical and biotechnology companies look for ways to shave years off research and development timelines and increase the chances of success amid rising costs. more More than 200 startups They are now competing to integrate AI directly into research workflows, attracting increasing interest from investors. Converge Bio It is the latest company to ride this transformation, securing new capital as competition intensifies in the field of AI-driven drug discovery.

The Boston and Tel Aviv-based startup, which helps pharmaceutical and biotechnology companies develop drugs faster using generative AI trained on molecular data, has raised $25 million in an oversubscribed Series A round, led by Bessemer Venture Partners. TLV Partners and Vintage Investment Partners also joined the round, along with additional support from unidentified executives at Meta, OpenAI, and Wiz.

In practice, Converge trains generative models on DNA, RNA, and protein sequences, then feeds them into pharmaceutical and biotech companies’ workflows to accelerate the drug development process.

“The drug development lifecycle has defined phases — from target identification and discovery to manufacturing, clinical trials and beyond — and in each of them, there are trials we can support,” Dov Gertz, CEO and co-founder of Converge Bio, said in an exclusive interview with TechCrunch. “Our platform continues to expand across these stages, helping bring new medicines to market faster.”

So far, Converge has rolled out customer-facing systems. The startup has already introduced three separate AI systems: one for antibody design, one for protein production optimization, and one for biomarker and target discovery.

“Take our antibody design system as an example. It’s not just one model. It consists of three integrated components. First, the generative model creates new antibodies. Next, predictive models filter those antibodies based on their molecular properties. Finally, the docking system, which uses a physics-based model, simulates 3D interactions between an antibody and its target,” Geertz continued. The value lies in the system as a whole, not in a single model, according to the CEO. “Our customers don’t have to put models together themselves. They get ready-to-use systems that plug directly into their workflow.”

The new financing comes about a year and a half after the company raised an amount $5.5 million seeds The tour is in 2024.

Since then, the two-year-old startup has expanded rapidly. Converge has signed 40 partnerships with pharmaceutical and biotech companies and currently runs about 40 programs on its platform, Gertz said.. It works with clients across the US, Canada, Europe, Israel and is now expanding into Asia.

The team has also grown quickly, growing to 34 employees from just nine in November 2024. Along the way, Converge began publishing public case studies. In one, the startup helped a partner increase protein production by 4 to 4.5X in a single computational iteration. In another case, the platform generated antibodies with extremely high binding, down to the nanomolar range, Gertz noted.

AI-based drug discovery is seeing a significant increase in interest. last yearEli Lilly has teamed up with Nvidia to build what the two companies call the pharmaceutical industry’s most powerful supercomputer for drug discovery. In October 2024, developers will be late Google DeepMind’s AlphaFold project won the Nobel Prize in chemistry to create AlphaFold, an artificial intelligence system that can predict protein structures.

When asked about the momentum and how it is shaping Converge Bio’s growth, Gertz said the company is seeing the largest financial opportunity in the history of life sciences and that the industry is shifting from “trial and error” approaches to data-driven molecular design.

“We feel deep momentum, especially in our inboxes,” Gertz told TechCrunch. “A year and a half ago, when we founded the company, there was a lot of skepticism.” He added that these doubts disappeared remarkably quickly, thanks to successful case studies from companies like Converge and from academia.

Large language models are gaining attention in drug discovery for their ability to analyze biological sequences and suggest new molecules, but challenges such as hallucinations and accuracy remain. “In text, hallucinations are usually easy to spot,” the CEO said. “In molecules, validation of a new compound can take weeks, so the cost is much higher.” To address this problem, Converge is integrating generative models with predictive models, filtering out novel molecules to reduce risk and improve outcomes for its partners. “This filtration process is not perfect, but it significantly reduces risk and delivers better results for our clients,” Gertz added.

TechCrunch also asked about experts like Yann LeCun, who have remained at the company Skeptical about using LLMs. “I’m a big fan of Yann LeCun, and I completely agree with him. We don’t rely on textual models for basic scientific understanding. To truly understand biology, models need to be trained on DNA, RNA, proteins, and small molecules,” Geertz explained.

Text-based LLMs are used only as support tools, for example, to help clients navigate the literature on generative molecules. “It’s not our core technology,” Gertz said. “We’re not tied to a single architecture. We use LLMs, diffusion models, traditional machine learning, and statistical methods when it makes sense.”

“Our vision is that every life sciences organization will use Converge Bio as a generative laboratory for AI,” Gertz said. “Wet labs will always exist, but they will be paired with generative labs that computationally create hypotheses and molecules. We want to be that generative laboratory for the entire industry.”