My undergraduate research was in a field known as molecular programming, where the goal is to program chemistry to behave in a desired manner (just like we program computers). I worked on tools for DNA sequence design—tools that can determine the sequence of base pairs a set of DNA molecules should have if we want the molecules to interact in a certain way. Thanks to Joseph Berleant, who continued and finished the project after I graduated, the resulting software package can be found on github here.
My graduate research was in machine learning theory. I worked on algorithms and theoretical guarantees for making use of unlabeled data: active and semi-supervised learning and domain adaptation. Much of this research involved looking at heuristics used by machine learning practitioners and trying to understand why and when they work well from the theoretical perspective.
Note: author names below are listed in alphabetical order.
At Oncora, we amassed a large dataset of medical record and treatment data, mostly around radiation therapy treatments for cancer. We used this data to train machine learning models that can predict the likelihood of various treatment outcomes. Some of our research focused purely on determining how well we can predict these outcomes, while other projects focused on usability and interpretability of the models through prediction-level confidence intervals and explanations. We even deployed some predictive models in the clinic at one of our partner sites and saw how the use of our models changed clinical practice in a meaningful way. Most of our numerous abstracts and journal articles can be found here.