Somashekhar Kulkarni

Topic: Benchmarking Bio-image analysis workflows on HPC and EC platforms - Setup image denoising workflows based on U-Net and image segmentation workflows based on traditional methods on HPC cluster and EC server - Benchmarked containerized Bio-image analysis workflows on HPC cluster and EC server - Statistically analyzed the benchmark results, using bootstrap-based confidence intervals - Devised a strategy to choose the optimal computing system for the execution of a particular workflow

- Literature survey on uncertainty techniques for neural networks, Bayesian ML, Gaussian Processes, and related topics - Understanding, implementation, and testing of Monte-Carlo Dropout and Ensembles techniques on synthetic data for regression-like tasks in PyTorch - Implementation of Gaussian Processes in MALA using GPyTorch and speedup with the help of GPUs - Setup a model consisting of two serial Gaussian Processes in MALA using GPyTorch to compute the total energy of a molecular system

Worked as a Student assistant at the Geometric modeling and visualization (GMV) research group which is a part of the Institute of Geometry at TU Dresden. I was involved in the design, production, and analysis of connectors for Concrete elements which are to be used in building large structures.

Involved in the installation and commissioning activities of a Vehicle testing cell at Horiba India Technical Centre, Pune, India

Involved in the modeling, meshing of cylindrical geometries similar to the TOSQAN test facility, and simulating the Spatio-temporal transport process related to stratification, natural circulation and containment wall condensation in these geometries using ANSYS CFX

- Involved in the design and development of rugged enclosures/chassis for electronic equipment based on the customer requirements - Entrusted with the task of performing structural & thermal simulations so that the products are in conformance with MIL-STD-810 - ESS (Environmental stress screening) involves checking the structural integrity of the chassis by simulating structural tests - Thermal simulations involve simulating the process of conduction cooling or convection cooling

Relevant Coursework: Deep Learning, Computer Vision, Machine Learning, High-Performance Computing, GPU Programming, Digitization and Data Analytics, Scientific Programming with C++, Scientific Programming with Java, Stochastics, Numerical methods, Advanced problem solving and search