Drylab raises $3 million to expand AI research agents
Drylab AI announced US$3 million in seed funding on October 13. An investor release says Vertex Ventures Southeast Asia & India led the round, with participation from PatSnap.
The capital will fund enterprise sales recruitment, engineering and biology hiring, and security and compliance certifications sought by large research organisations, according to the announcement.
Founded in 2025, Drylab launched commercially in February 2026. It says paying customers include pharmaceutical companies, cancer centres, research institutes and universities in the United States, Europe and Asia.
How the research agents work
Drylab targets the computational side of biomedical research. Its documentation describes a workspace that combines a conversational interface, a notebook where generated code runs, and persistent file storage called Vault. Supported work includes genomics, imaging, molecular modelling and statistical analysis. These capabilities concern data and software workflows rather than physical laboratory automation.
Its analysis guide separates routine work, such as plots and statistical tests, from more involved discovery tasks and complex pipelines. Requests begin in plain language, but the resulting code runs in a notebook and reports back through the conversation.
The product also leaves decisions with scientists. Its analysis guide describes editable plans, clarification questions and pauses for scientific choices. Users can approve individual steps or enable automatic plan approval. That makes the level of supervision configurable, an important qualification to broad descriptions of agents carrying out research independently.
Private deployment has several forms
Drylab says its enterprise platform can place data, computation and agents inside a customer’s infrastructure. Organisations can use their own model providers or self hosted models, and connect existing storage and pipelines through APIs. The company also advertises installations that are fully disconnected from outside networks. These are deployment claims, not independently verified security findings.
The desktop app follows a different arrangement. A technical post says source files remain local by default and are mounted with read only access. It also identifies an exception for code execution output sent to AI agents, alongside optional cloud computing for larger jobs. Keeping the original dataset on a workstation therefore does not establish that every result stays there.
What buyers can check today
The public plan comparison lists on premises deployment and private integrations as enterprise features. It labels autonomous discovery and model training as beta or proof of concept offerings in several paid tiers. The advertised capabilities should therefore be checked against the specific plan and deployment being purchased.
The announcement does not disclose a valuation, customer count, measured productivity gains or the financing closing date. Those gaps leave the scale of Drylab’s commercial traction difficult to assess.
Illustrative research photo by National Cancer Institute on Unsplash. Used under the Unsplash License.
