BioRender has unveiled an AI agent purpose-built for generating scientific figures.
The news: The Toronto software startup announced the launch of Leo, the latest addition to its suite of AI tools, on Tuesday. A BioRender spokesperson told BetaKit over email that while researchers could previously prompt the companyâs platform to produce scientific figures, Leo now provides them with expert guidance throughout the generation process as they plan, sketch, and build scientific images in the same fashion as âa trained medical illustrator.â
Leo uses the best available AI models for each step, while leveraging a library of scientist-vetted icons and templates. The company claims Leo can help clients produce accurate, clear, and editable figures with greater fidelity than general-purpose AI. This is important for a startup that has set its sights on addressing a growing issue: âthe erosion of trust in science caused by a proliferation of AI-generated images.â
Leo is part of BioRenderâs expanding suite of AI-native tools. A BioRender spokesperson told BetaKit over email that while users could previously prompt BioRender to produce scientific figures, Leo now provides them with expert guidance throughout the generation process. Making these figures requires more than just understanding the science, and Leo uses the best available AI models for each step, while leveraging a library of scientist-vetted icons and templates.
From the source: âNearly every week, a new AI image-generation tool is released,â BioRender co-founder and CEO Shiz Aoki said in a statement. âBut scientists canât rely on them because they treat scientific figures the way they would any other type of illustration. Leo is designed differently. It doesnât jump from prompt to picture. Instead, it provides scientists with actual figure-making expertise, asking questions, offering sketches, and building images that a researcher can continue to refine by hand.â
The context: Aoki previously worked as lead science illustrator at National Geographic. She teamed up with repeat tech entrepreneur Ryan Marien and software engineer Katya Shteyn (the companyâs CTO) to found BioRender in 2017. Since then, the company has been working to standardize the visual language of biology through tech by helping researchers produce scientific graphics in minutes rather than hours or days.
Aoki noted that the deterministic nature of science, which leaves âno room for randomness or chance,â makes it particularly difficult to visualize. While general-purpose text-to-image generative AI tools such as ChatGPT, DALL-E, Gemini, and Midjourney can produce scientific research images, they often fail to meet academic and publication standards because they were not trained on specialized domain knowledge. Given this, they can be more prone to hallucination and require significant re-prompting and high token spending to produce useful illustrations.
Final thought: BioRender has built a so-called AI-powered âCanva for Scienceâ that helps more than four million scientists, thousands of research institutions, all of the worldâs top 30 pharmaceutical firms, and 15 Nobel Prize-winning labs draw biological diagrams more easily and communicate their research more efficiently. The profitable, 200-person startupâs tech has even caught the attention of American AI giant Anthropic, which struck a deal last October to integrate BioRenderâs library into its Claude for Life Sciences model. BioRender parlayed that progress into a December 2025 secondary investment from biotech investor Dimension Capital that valued the startup at $900 million USD.
UPDATE (10/07/26): This story has been updated to include additional context and information provided by BioRender.
Feature image courtesy Pexels. Photo by Polina Tankilevitch.