Our Investment in Phylo: Building the Intelligence Layer for Biology
back to All
The next great frontier for intelligence is science — and no field is more consequential than biology. Its data is abundant, its results can be verified through analysis and experiment, and its discoveries — better drugs, diagnostics, and therapies — create extraordinary value for patients and society. Yet biology still cannot iterate at the speed of software.
Today, Phylo announced its $13.5 million seed round, we are proud to back Kexin Huang, Yuanhao Qu, and the Phylo team as they build the intelligence layer for biology.
Biology has the tools, but not the workflow
Coding was the first domain frontier models transformed, because its feedback loop is tight: generate, run, test, improve. Biology's is fragmented. Literature, experimental data, analysis software, scientific databases, and lab protocols all live in separate systems, and scientists lose hours finding resources, moving data between tools, debugging pipelines, and coordinating handoffs. Biology doesn't lack powerful tools; it lacks a unified way to use them.
That is exactly what makes it suited to agents. An agent can hold context across a project, choose the right tools, run multi-step analyses, inspect intermediate results, and adapt when the workflow changes — much closer to how a scientist actually works than a single model call.
The intelligence layer for biology
AI for Science has two layers. The creation layer designs molecules, proteins, and therapeutic candidates. The intelligence layer coordinates models, data, software, and experiments around a scientific objective. Phylo is building the second.
Its flagship product, Biomni Lab, is an Integrated Biology Environment — one workspace where researchers direct agents through rigorous, large-scale scientific work. More than 300 databases, software systems, and analytical tools are integrated into a single environment, so a scientist can move from literature and data analysis to model building and experimental design without rebuilding the workflow or losing context. Its use cases already span RNA sequencing, biomarker design, protein structure prediction, and clinical-trial landscaping; in a collaboration with Ginkgo Bioworks, Biomni Lab compressed more than ten complex cell-painting and transcriptomic analyses from weeks into hours.
This is not another chatbot for science. The scientist sets direction and exercises judgment; the agents handle execution across a growing universe of tools and data.
Why Phylo
Three things gave us conviction.
- The right product abstraction. Like an IDE for software, the Integrated Biology Environment collapses the friction between an idea and its execution.
- A compounding technical path. Phylo pairs frontier models with real scientific workflows, open research, and real-world evaluation — so every use teaches the team where agents succeed, where they fail, and how to make them more capable, steerable, and reliable.
- A rare cross-domain team — and the reason we kept leaning in.
Kexin has spent years working toward autonomous AI scientists, long before the category had a name; Yuanhao brings deep experimental and computational biology expertise, including agentic systems for CRISPR research. Together they bridge Stanford computer science and medicine — a combination that is genuinely rare. In our conversations, what stood out was not just their pedigree but their clarity about how science actually happens, and an unusual ability to ship. Everyone in our network who has worked with them told us a version of the same thing: academic depth paired with real execution. It is among the most impressive founding teams we have backed.
The ambition is bigger than automating tasks. If Phylo succeeds, scientists will spend less time operating tools and more time choosing the right questions, interpreting evidence, and designing the next experiment — more hypotheses tested, shorter cycles, discoveries that compound faster. We believe the most important AI for Science companies will understand both frontier intelligence and how science actually happens. Phylo is one of them — and we are grateful to build the future of biology alongside Kexin, Yuanhao, and the team.
Our Investment in Phylo: Building the Intelligence Layer for Biology
Back to All
The next great frontier for intelligence is science — and no field is more consequential than biology. Its data is abundant, its results can be verified through analysis and experiment, and its discoveries — better drugs, diagnostics, and therapies — create extraordinary value for patients and society. Yet biology still cannot iterate at the speed of software.
Today, Phylo announced its $13.5 million seed round, we are proud to back Kexin Huang, Yuanhao Qu, and the Phylo team as they build the intelligence layer for biology.
Biology has the tools, but not the workflow
Coding was the first domain frontier models transformed, because its feedback loop is tight: generate, run, test, improve. Biology's is fragmented. Literature, experimental data, analysis software, scientific databases, and lab protocols all live in separate systems, and scientists lose hours finding resources, moving data between tools, debugging pipelines, and coordinating handoffs. Biology doesn't lack powerful tools; it lacks a unified way to use them.
That is exactly what makes it suited to agents. An agent can hold context across a project, choose the right tools, run multi-step analyses, inspect intermediate results, and adapt when the workflow changes — much closer to how a scientist actually works than a single model call.
The intelligence layer for biology
AI for Science has two layers. The creation layer designs molecules, proteins, and therapeutic candidates. The intelligence layer coordinates models, data, software, and experiments around a scientific objective. Phylo is building the second.
Its flagship product, Biomni Lab, is an Integrated Biology Environment — one workspace where researchers direct agents through rigorous, large-scale scientific work. More than 300 databases, software systems, and analytical tools are integrated into a single environment, so a scientist can move from literature and data analysis to model building and experimental design without rebuilding the workflow or losing context. Its use cases already span RNA sequencing, biomarker design, protein structure prediction, and clinical-trial landscaping; in a collaboration with Ginkgo Bioworks, Biomni Lab compressed more than ten complex cell-painting and transcriptomic analyses from weeks into hours.
This is not another chatbot for science. The scientist sets direction and exercises judgment; the agents handle execution across a growing universe of tools and data.
Why Phylo
Three things gave us conviction.
- The right product abstraction. Like an IDE for software, the Integrated Biology Environment collapses the friction between an idea and its execution.
- A compounding technical path. Phylo pairs frontier models with real scientific workflows, open research, and real-world evaluation — so every use teaches the team where agents succeed, where they fail, and how to make them more capable, steerable, and reliable.
- A rare cross-domain team — and the reason we kept leaning in.
Kexin has spent years working toward autonomous AI scientists, long before the category had a name; Yuanhao brings deep experimental and computational biology expertise, including agentic systems for CRISPR research. Together they bridge Stanford computer science and medicine — a combination that is genuinely rare. In our conversations, what stood out was not just their pedigree but their clarity about how science actually happens, and an unusual ability to ship. Everyone in our network who has worked with them told us a version of the same thing: academic depth paired with real execution. It is among the most impressive founding teams we have backed.
The ambition is bigger than automating tasks. If Phylo succeeds, scientists will spend less time operating tools and more time choosing the right questions, interpreting evidence, and designing the next experiment — more hypotheses tested, shorter cycles, discoveries that compound faster. We believe the most important AI for Science companies will understand both frontier intelligence and how science actually happens. Phylo is one of them — and we are grateful to build the future of biology alongside Kexin, Yuanhao, and the team.
Our Investment in Phylo: Building the Intelligence Layer for Biology
Back to All
The next great frontier for intelligence is science — and no field is more consequential than biology. Its data is abundant, its results can be verified through analysis and experiment, and its discoveries — better drugs, diagnostics, and therapies — create extraordinary value for patients and society. Yet biology still cannot iterate at the speed of software.
Today, Phylo announced its $13.5 million seed round, we are proud to back Kexin Huang, Yuanhao Qu, and the Phylo team as they build the intelligence layer for biology.
Biology has the tools, but not the workflow
Coding was the first domain frontier models transformed, because its feedback loop is tight: generate, run, test, improve. Biology's is fragmented. Literature, experimental data, analysis software, scientific databases, and lab protocols all live in separate systems, and scientists lose hours finding resources, moving data between tools, debugging pipelines, and coordinating handoffs. Biology doesn't lack powerful tools; it lacks a unified way to use them.
That is exactly what makes it suited to agents. An agent can hold context across a project, choose the right tools, run multi-step analyses, inspect intermediate results, and adapt when the workflow changes — much closer to how a scientist actually works than a single model call.
The intelligence layer for biology
AI for Science has two layers. The creation layer designs molecules, proteins, and therapeutic candidates. The intelligence layer coordinates models, data, software, and experiments around a scientific objective. Phylo is building the second.
Its flagship product, Biomni Lab, is an Integrated Biology Environment — one workspace where researchers direct agents through rigorous, large-scale scientific work. More than 300 databases, software systems, and analytical tools are integrated into a single environment, so a scientist can move from literature and data analysis to model building and experimental design without rebuilding the workflow or losing context. Its use cases already span RNA sequencing, biomarker design, protein structure prediction, and clinical-trial landscaping; in a collaboration with Ginkgo Bioworks, Biomni Lab compressed more than ten complex cell-painting and transcriptomic analyses from weeks into hours.
This is not another chatbot for science. The scientist sets direction and exercises judgment; the agents handle execution across a growing universe of tools and data.
Why Phylo
Three things gave us conviction.
- The right product abstraction. Like an IDE for software, the Integrated Biology Environment collapses the friction between an idea and its execution.
- A compounding technical path. Phylo pairs frontier models with real scientific workflows, open research, and real-world evaluation — so every use teaches the team where agents succeed, where they fail, and how to make them more capable, steerable, and reliable.
- A rare cross-domain team — and the reason we kept leaning in.
Kexin has spent years working toward autonomous AI scientists, long before the category had a name; Yuanhao brings deep experimental and computational biology expertise, including agentic systems for CRISPR research. Together they bridge Stanford computer science and medicine — a combination that is genuinely rare. In our conversations, what stood out was not just their pedigree but their clarity about how science actually happens, and an unusual ability to ship. Everyone in our network who has worked with them told us a version of the same thing: academic depth paired with real execution. It is among the most impressive founding teams we have backed.
The ambition is bigger than automating tasks. If Phylo succeeds, scientists will spend less time operating tools and more time choosing the right questions, interpreting evidence, and designing the next experiment — more hypotheses tested, shorter cycles, discoveries that compound faster. We believe the most important AI for Science companies will understand both frontier intelligence and how science actually happens. Phylo is one of them — and we are grateful to build the future of biology alongside Kexin, Yuanhao, and the team.