All Posts

Accelerating biotech workflows by combining metabolic modeling (CarveMe/COBRApy) with SMAC3 Bayesian optimization to simulate and discover ideal fungal growth environments.

Built BioX, a web-based digital twin that automates fungal media optimization using CarveMe, COBRApy, and SMAC3 Bayesian optimization.

How an adaptive reward shifting mechanism overcomes conservative AI behavior to unlock unprecedented sample efficiency in Dynamic Algorithm Configuration.

Our paper on reward design for RL-based Dynamic Algorithm Configuration won the Best Paper Award at GECCO 2025 in Málaga, Spain.

How integrating 2D molecular graphs with ensembles of 3D conformers via an efficient, geometry-aware barycenter pooling mechanism yields state-of-the-art molecular property prediction.

How combining a dynamic two-stage uncertainty frame with a localization-heavy Emphasis loss breaks the data-hungry bottleneck of deep object detectors.