Ian Mackey

Lunar super-resolution via normalizing flows to improve feature detection while maintaining uncertainty quantification. Leverages conditional blocks to enable a continuous scale factor.

End-to-end pipeline for resident space object shape modeling. Specifically responsible for the generative query network used for implicit scene understanding.

Evaluated three different methods of identifying common medical landmarks in chest CT scans. Personally responsible for encoder network using elastic graph deformation to register landmarks similarly to how facial recognition is commonly acheived.

Implemented a cuda-enabled gym-like environment and proximal policy optimization variant to balance a portfolio given only three years of historical data. Currently working towards multi-task and continual learning approach to generalize current results.