I’m a machine learning leader and scientist. I most recently founded and led the machine learning organization at Terray Therapeutics, building generative, agentic, and predictive systems for high-throughput drug discovery. My team’s work spanned multimodal equivariant transformers, latent diffusion for molecular design, and transferable pretrained models for co-posing and potency, integrated into a fully autonomous design–make–test–analyze loop.

John Parkhill

Background

My path to ML ran through the physical sciences. I earned a B.S. in Mathematics and Chemistry from the University of Chicago (2005) and a Ph.D. in Quantum Chemistry from UC Berkeley (2010), working with Martin Head-Gordon. I then completed a postdoc with Alán Aspuru-Guzik at Harvard, before joining the faculty at the University of Notre Dame as an Assistant Professor, where I developed the TensorMol neural network model chemistry and was awarded an NSF CAREER grant.

From there I spent several years as Head of Quantitative Research at Artemis Capital Management, leading research and data operations for a long-volatility macro fund and building deep-learning systems for volatility modeling and risk. More recently I returned to science, this time on the applied side, to build the ML function at Terray Therapeutics.

Contact

This site hosts occasional notes on machine learning, scientific computing, and quantitative modeling.