Convoke Raises $8.6M to Build the AI Operating System for Biopharma

Today, an entrepreneur can start a software startup with little else apart from an idea and a laptop. The vibrant ecosystem of software startups is a direct result of decades of progress in infrastructure (e.g. the internet, cloud services, open source software) that have abstracted away large parts of the complexity of running a business. This has made it possible to focus on solving specific problems and create immense value at low cost by combining and extending existing technologies.

This dynamic is absent in the pharmaceutical industry: the costs, complexity, and technology required to develop a new drug are immense—and rising. Nearly two years ago one of our founders wrote a history of Paul Janssen, a Belgian physician who founded what would become one of the world’s largest pharmaceutical companies when he was just 27. He set up his fledgling company on a spare floor in his parent’s firm, with a loan of 50,000 Belgian francs (~$12,000 today). It would be effectively impossible to start a biotech company today with such a paltry sum. Yet, Janssen succeeded.

When Janssen started his company, it was relatively cheap and easy to find and test new ideas for drugs. Unlike computer science, biotechnology is a physical science, and like many other physical industries costs have risen dramatically over time as a function of scarce inputs, particularly labor. For example, clinical trials now cost as much as $100,000 per patient to run, and sometimes up to $300,000 or even $500,000 per patient for complex designs—much of this is wages for hospital staff and administrators. 

Worsening matters, over the years, the easy ideas were mined and new regulation was layered on. Experiments have become larger and more complex, as we are forced to more deeply plumb the depths of biology for new insights into disease. Even if clinical trials were made cheaper, there aren’t enough patients in the world to test all our ideas for new treatments in a practicable timeframe.

If we ever hope to cure all disease we must find a way to escape the constraints imposed by the physical world—we need to move as much work as we can in silico.

Our best hope so far to make this transition is through artificial intelligence (AI), and we’re seeing progress along two parallel tracks:

• The first revolution is generative drug design; tools like AlphaFold have unlocked new capabilities for drug discovery that drastically shorten design and test cycles and time from concept to plausible drug candidate

• The second, and perhaps more neglected revolution, is in drug development infrastructure. Between an idea and a patient receiving treatment is a morass of knowledge work: regulatory submissions, clinical protocols, investigator brochures, and manuscripts—thousands of pages of documentation, each requiring deep expertise and months of work. Large language models (LLMs) in particular promise to greatly accelerate the creation of these regulatory and decision-making artifacts

If we are to benefit from a wave of new drugs promised by better discovery tools, we need more efficient development infrastructure. And today we’re announcing our $8.6M seed financing from Kleiner Perkins, Dimension Capital, and others to work towards that future (you can read about the fundraise in Axios here).

Today, software startup founders don't have to spin up their own server cluster, build payment processing from scratch, write their own authentication system, manage physical databases, or spend months setting up deployment infrastructure. Cloud platforms, APIs, and open-source tools have abstracted away these complexities that once required massive capital and specialized expertise. Our hope is that we can do the same for drug development knowledge work. 

Our team combines top technical talent from Google and Applied Intuition with experts from every part of the drug development cycle—from scientific discovery to the biggest commercial launches of the past decade. We've seen similar transformations before; several of us witnessed the automotive industry's transition to autonomy firsthand. We know what it takes to build systems that don't just assist but eventually replace entire workflows.

Today, we’re working closely with leading biotech and pharma companies to identify their most critical bottlenecks and build AI workflows that eliminate them. What begins as AI assistance—helping implement an edit in a regulatory submission or conducting a literature review—will evolve into complete autonomous workflows. Our infrastructure already allows customers to task tens of thousands of digital colleagues (agents) with traversing vast landscapes of scientific and medical knowledge to find answers and create reports that would otherwise take months of graft.

It’s early, but we’re seeing signs that our technology can help prioritize lab or clinical work, and increase the speed at which teams can iterate on an idea or a business before they jump into expensive physical testing. One day, we hope that the infrastructure we’re building will collect and synthesize information from the entire landscape of sources, and autonomously guide new drugs from an idea, through testing, regulatory approval, and commercial launch. This process will generate data streams that can feed back into, and improve, the next iteration of their designs.

If we’re successful, we will democratize access to expertise and shave years off of drug development timelines for the largest and the smallest organizations in the industry. 

If you're interested in contributing to this mission, reach out at founders@convoke.bio.