Rakesh Halder Email & Phone Number

New York, Ny, Us

Barcelona, Catalonia, Spain

Ann Arbor, Michigan, United States

• Research involves developing low-dimensional surrogate physics models (reduced-order models) using deep learning, machine learning, and optimization.• Utilize convolutional autoencoders to provide low-dimensional solution manifolds of nonlinear partial differential equations (PDEs) to build highly accurate and efficient reduced-order models.• Train variational autoencoders (VAEs) using custom annealing schedules to control the structure of latent variables for enhanced interpretability and prevention of posterior collapse.• Work on large-scale adjoint optimization methods to approximate high-fidelity fluid dynamics simulations with lower-fidelity turbulence models using high-performance computing (HPC) clusters.• Optimize the training data for data-driven and physics-based reduced-order models using Isomap and Gaussian process regression.• Submitted a successful research proposal to a National Science Foundation (NSF) program for 1.2M core-hours of HPC allocation.

Ann Arbor, Michigan, United States

Graduate Student Instructor for AEROSP 423: Computational Methods for Aerospace Engineering

Toronto, Ontario, Canada

• Conducted research on AI-driven surrogate models for time-dependent fluid dynamics simulations.• Used convolutional autoencoders to develop low-dimensional time-series models of high-dimensional dynamical systems.• Combined multiple data sets to develop an accurate autoregressive forecasting framework using recurrent neural network ensembles that generalized to unseen multivariate time-series data over long prediction horizons.• Compared the performance of using ensemble LSTMs to transformers for time-series prediction.• Effectively reduced the cost of high-fidelity simulation by up to 85% while retaining high levels of accuracy.

New Brunswick, New Jersey, United States

• Investigated the transport of a rigid particle in multi-phase flow with a wetting and non-wetting fluid using molecular dynamics simulations.• Submitted honors thesis through the James J. Slade Scholars Program at Rutgers School of Engineering.

West Lafayette, Indiana, United States

Worked on a computational model to investigate the impact of manufacturing errors in spur gears on the volumetric efficiency of external gear pumps.

Raleigh, North Carolina, United States

Doctor Of Philosophy - Phd, Aerospace Engineering And Scientific Computing

Bachelor’S Degree, Mechanical Engineering

Activities and Societies: Rutgers University Engineering Honors Academy, Tau Beta PiMinors in Mathematics and Physics