• Lead the creation of differentiable CFD frameworks in JAX, extending capabilities to complex geometries and non-periodic boundary conditions using immersed boundary methods • Design, implement, and containerize code (Docker/enroot) to ensure reproducibility and scalability across multi-node GPU HPC environments • Develop optimized parallel linear solvers (FFT or matrix decompositions) and gradient-based scripts to iteratively modify reactor designs based on catalyst activity • Rigorously validate simulation accuracy against benchmark results and fixed-bed reactor configurations for both non-reactive and reactive flow regimes • Partner with AI model developers to integrate property predictions and generate comprehensive technical reports summarizing algorithm scalability and differentiability