I believe open, reproducible research accelerates discovery. Below are the tools and frameworks I have developed and released publicly. Each project page will grow into a technical blog with tutorials, notes, and updates.

DeepBubbleVelocimetry

Stars   Deep Learning Optical Flow Multiphase Flow

GitHub

CNN-based (PWC-Net optical flow) velocity measurement tool purpose-built for bubbly two-phase flows. The first deep-learning flow-diagnostic tool capable of handling extreme bubble densities (up to 58% void fraction) — far beyond the 11% limit of conventional PTV methods. Includes pre-trained model weights and a synthetic bubble image generator for fine-tuning.

📄 Related paper: Choi*, Kim* & Park, Sci. Rep., 12, 11879 (2022)

Open-source FSI Solver

OpenFOAM CalculiX preCICE Fluid-Structure Interaction

arXiv

Fully reproducible 3D FSI simulation framework coupling OpenFOAM (CFD), CalculiX (FEA), and preCICE (coupling library) for compliant nozzle dynamics. Deployed in Singh & Choi et al. (2026) to study elastic wave propagation in flexible nozzles and its role in impulse enhancement. Full solver and case files will be publicly released with the paper.

📄 Related paper: Singh, Choi, Bhamla & Bose, arXiv:2605.17319 (2026). Under review.

Deep-learning Bayesian Surrogate for Nozzle Optimization In preparation

Bayesian Optimization Surrogate Modeling PyTorch Propulsion

A deep-learning Bayesian surrogate model for soft nozzle thrust optimization. Combines PIV/dynamometry experimental data with Gaussian process regression and neural network surrogate models to efficiently explore the nozzle geometry design space without exhaustive experiments. To be released alongside the manuscript.