Justin Miller - Portfolio of Work
Projects completed during the DECES Sc.M. program at Brown University.
Accelerating the Transient Conduction Heat Transfer Problem w/ GPUs
Developed solver for wire temperature in varying weather conditions using implicit finite difference.
Implemented shared and distributed memory parallelization approaches on a time-space grid using C++, CUDA and MPI to reduce a 1hr Python simulation to a 4 second runtime (when using 4 GPUs).
Modeling Inverse Heat Transfer using Physics-Informed Neural Networks
Leveraged PINNs to solve inverse heat transfer in a flow around a cylinder with patchy sensor data.
Encoded physical behavior with unknown boundary conditions by incorporating governing PDEs via network loss
Supercompressible Meta-Material Design using Machine Learning
Examining the predictive strength of various algorithms on ABAQUS FEA sims.
Used SVM, Gaussian Process Regression and NNs to analyze simulated output and generate property
3D Line Reconstruction: Reviewing LIMAP and CLMAP
Modified LIMAP and CLMAP R&D repositories for practical use on raw video datasets.
Examined scenes from Tanks & Temples dataset, identifying weaknesses and profiling strengths in geometric identification.