Two research fronts, one common thread.
My doctorate (FGV/EMAp) studies stochastic optimization for the Brazilian power system: SDDP variants, cut selection, and everything that the national operator runs on. My master's (COPPE/UFRJ) traces modern generative models through the lens of Optimal Transport, from normalizing flows to score-based diffusion to flow matching, using the through-line idea that is in both: moving mass from where it is to where it should be.
The animation above is the literal subject of the master's thesis: a forward diffusion that melts the word into a hot cloud, then a reverse process that reassembles it.
Data-Driven Diffusion-Based Super-Resolution for Improvement of Reduced-Order Model Predictions in Fluid Dynamics
My first paper as lead author — a diffusion model that sharpens reduced-order fluid-dynamics predictions, faster than running the full simulation!
Improving accuracy of parametric surrogate model for turbidity currents using diffusion-based super-resolution
Diffusion-based super-resolution applied to turbidity-current surrogate models — recovering fine structure that the cheap solver loses.
Boundary Conditions for Hydrothermal Operation Planning: The Infinite-Horizon Approach
How to close the end of planning horizon problem — infinite-horizon boundary conditions for the Brazilian hydrothermal dispatch problem.
soon™ there will be more.