AI is transforming drug discovery, with a major emphasis on target identification and molecular design. That work matters and hopefully will lead to a large number of safe molecules arriving in the clinic. We choose to work on what happens next.
Dizzaroo is about development: the translational and clinical work that takes a molecule and makes it a drug.
I am a physician-scientist. I have spent eighteen years in drug development, in small biotechs short on cash and in large pharma, and then two years working hands-on with AI. Most recently I led clinical development of Enhertu for five years and spent five years with Pfizer before that. Large pharma is known for being good at development, and I have watched closely how it is done. However, the same three gaps appear everywhere, whatever the budget.
I could not know everything I needed to know. Designing one trial means weighing thousands of factors: what FDA has said before, the current guidance, the standard of care, access, cost, the new therapies that could catch you by surprise. Nearly all of that information is available. It is scattered, slow to retrieve, and too much for any person to reason over. I advised other teams on other breast cancer programs, and I still could not say that I knew everything there was to know about the program I was leading, or that I could bring it to bear at the right moment, every time. Nobody can.
A large body of the work after a decision is mechanical. A large study runs across many countries, hundreds of sites, hundreds of people, and several regulators. Everything has to be written down and done in order. Starting one study takes thirty to thirty-five documents. It is tedious and slow, and experienced people spend their days on it.
The work breaks at the handovers. This is the biggest source of quality problems and lost time. Work passes between functions, between organizations, and between old systems that do not talk to each other. Something is lost or misread at every pass.
The gaps open before the clinic, too. Which molecule to take forward, and for which patients, is decided with what tissue, imaging and biology can tell you. That translational work usually sits apart from the clinical plan it is meant to inform. We treat the two as one piece of work.
A leader I worked for used to say that quality comes first, then time, then resources. When you optimize, you optimize in that order. I say this with humility:
For the first time, all three can improve together.
Quality begins with the decision. When every relevant fact is used and reasoned over, you decide better. Fewer amendments, less rework and less lost time follow from that.
Developing a drug takes thousands of decisions, made by many intelligent agents. They have raw intelligence. We are building what makes them deliver: the harness, the system around the models, for clinical development.
We are also building the organization that uses it, an AI-native development organization, where drug developers and engineers work one way, on one platform.
Our first job is to do today's development far better. Next is to work out what clinical development looks like beyond the traditional trial. That is for another day.
We have been at this quietly for nearly two years, and it is no longer hypothetical. We will show more soon. If you are developing a medicine and want to work this way, I would love to talk.
Dhiraj Gambhire, MD, MPHFounder and CEO, DizzarooLinkedIn