中文标题待翻译

The cost of sequencing a human genome fell from about US$95 million in 2001 to about US$500 by 2022, according to the US National Human Genome Research Institute. Biology became a data business almost overnight. The ability to make sense of that data did not keep pace.

Most laboratories now generate more sequencing, imaging and assay data than their staff can analyse. The work runs through thousands of specialised bioinformatics tools (software that turns raw biological data into something a scientist can interpret), patchy computing setups and a chronic shortage of people who can operate both. Scientists can spend weeks coordinating analyses and fixing broken pipelines before they get an answer.

Also Read: How Vietnam is emerging as a leading AI builder ecosystem in Southeast Asia

Drylab AI thinks software agents can do much of that waiting for them. The Singapore-headquartered startup, founded by a Vietnamese team, has raised US$3 million in seed funding led by Vertex Ventures Southeast Asia & India. PatSnap, the AI-driven innovation intelligence company, also participated, and will lend Drylab its life sciences network, data and AI capabilities to support its push into pharmaceutical and biotech R&D.

The money will go into enterprise sales hires, bigger engineering and biology teams, and the security and compliance certifications that large research organisations demand before they let any vendor near their data.

An agent that owns the whole workflow

Drylab’s pitch goes beyond the familiar AI copilot that waits for instructions. Its agents are meant to own an entire computational research workflow. They interpret the scientific objective, plan the analysis and pick from more than 800 integrated tools and databases. They also provision computing resources, run the analysis, interpret the results and troubleshoot failures without a human prompting each step.

The company’s differentiator is a proprietary verification model. It is trained on real scientific analysis runs, checks the agents’ analytical decisions and flags methodological errors so the agents can correct themselves. The design targets the main weakness of AI in science: a model that sounds confident while doing the statistics wrong.

That matters because trust remains the sticking point for agentic AI in the region. A recent study found Southeast Asian tech leaders are learning to trust AI agents, but not with production. In Singapore, firms are adopting agentic AI while audit trails remain thin. In drug discovery, where an analytical error can send a programme down the wrong path for years, scientists will be more cautious still.

“Realising this potential requires a framework that makes AI reliable, reproducible and secure within each organisation’s own environment,” said John Phan, co-founder and CEO.

The data never leaves the building

The second half of the pitch concerns who owns what the AI learns. Drylab can be deployed on a customer’s private infrastructure or even on local workstations. The platform improves with every analysis it runs, and that accumulated learning stays with the customer.

For pharmaceutical companies, whose most valuable asset is proprietary data, this is close to a precondition rather than a perk. It does, however, raise a question the announcement leaves unanswered. If each customer’s learning stays locked inside its own walls, how does Drylab’s core product get better across its customer base? The answer will shape whether Drylab compounds like a platform or scales like a consultancy.

Also Read: The hidden economics of autonomous AI agents

The company launched commercially only in February 2026. It says it has since signed paying customers among leading pharmaceutical companies, cancer centres, biomedical research institutes and universities in the US, Europe and Asia. None are named, and Drylab has not disclosed revenue or customer numbers. Eight months is a short track record, and enterprise pilots in pharma often take longer than that to become real contracts.

A Vietnamese team, a Singapore base, a global market

The founding story is unusually concrete. Drylab was set up in 2025 by CEO John Phan, who did AI research with ByteDance’s Seed team, and CTO Huy Phan, a former visiting research engineer at Andrew Ng’s Landing AI. The third co-founder is Chief Science Officer Nhi Nguyen, a PhD candidate in biomedical science at Nanyang Technological University. She taught herself bioinformatics to analyse her own lab’s data, and that experience led to Drylab.

Vertex’s Genping Liu framed the deal as proof of “global ambition coming out of Vietnam’s world-class tech talent pool”. The bet fits a broader trend: Vietnam’s AI funding has grown 13 times in two years. The country is also producing other AI-for-science plays, including N2TP, which Touchstone backed to build AI infrastructure for scientific discovery.

The investor mix deserves a closer look. Vertex is a long-time backer of PatSnap, so the strategic investor here is also a Vertex portfolio company. That could be a well-matched alliance, since PatSnap has spent years building its own in-house AI models for R&D intelligence and has a sales network that reaches into pharma. It could also be a tidy network effect within one fund. Customers will judge it by whether PatSnap’s channel actually opens doors.

Up against much bigger money

Drylab is entering a field where its rivals carry far heavier funding. In the US, Flagship Pioneering-backed Lila Sciences has raised hundreds of millions of dollars to build “scientific superintelligence”. The Eric Schmidt-backed FutureHouse is developing AI agents for research. Google has shown off an AI “co-scientist”, and Stanford researchers have open-sourced Biomni, a general-purpose biomedical agent.

Closer to home, India’s Elucidata sells a data platform that cleans and structures biomedical data for pharma customers. Workflow players such as Seqera, the company behind the widely used Nextflow pipeline language, already sit inside many bioinformatics teams.

Against that backdrop, US$3 million is a modest seed cheque. Drylab’s likely edge is focus: secure, locally deployed agents built for regulated labs, sold by a team that knows the pain from the bench. As one e27 contributor put it, AI is not the beginning of drug discovery but the accelerator, and it still needs good science underneath.

Also Read: The 27 SEA biotech firms betting on cells, fermentation, and code

Phan says the goal is to help scientists “move from biological questions to validated discoveries faster”. The next 18 months will show whether the word “validated” holds up. Drylab needs named customers, published benchmarks for its verification model, and independent evidence that its agents catch errors rather than introduce them.

The post Vietnamese-founded Drylab raises US$3M to put autonomous AI scientists inside pharma labs appeared first on e27.