"Unraveling complex patterns from data is key to advance biomedical discovery. Modern AI offers the tools to accomplish that."
I trained as a statistical physicist researching emergent behaviors for many body systems. I did my post-doctoral training at the Carl Woese Institute of Genomics and MIT, where I analyzed different kinds of biological data, combining conventional statistics with mathematical and predictive modelling. In 2018 I joined the Broad Institute of MIT and Harvard to lead our deep learning effort for building the Human Cell Atlas. In 2021, I moved to Genentech to run the AI/ML department in Research Biology. In my parallel life, I pursue my interests in writing open source software, cybersecurity, and developing the Linux system.
Post-doctoral years are a unique time where a scientist has enough experience to develop their own direction without the responsibility of a group leader. These transitional years can be challenging. But with the proper mentorship and guidance, one can learn how to develop their ideas into a mature research program. As a dedicated mentor, I commit a substantial part of my time to help my postdocs identify key questions in drug development where they can develop cutting-edge algorithms. Our team has deep expertise and is embedded within the leading drug development community.
biorxiv 272831v3 (2020)
My department broadly uses AI to enable and facilitate target discovery across Research Biology. Such an endeavor spans data types and scientific questions. Current directions range from conventional semantic segmentation of histopathology to learning disentangled representation of multi-modal data. A unique feature of doing AI within Genentech is that the goal is to gain scientific insight rather than meet engineering specifications. Therefore, we pay particular attention to interpretability and causal inference and also carry out fundamental research in these fields.