I started as a software engineer in March 2020 initially focusing on optimizations of machine learning workloads. This included research and implementation of mixed precision training and inference for Tensorflow models and working on stochastic optimization algorithms using MLIR/LLVM for search space code generations.I eventually transitioned to frontend/middle end compiler engineer for a hardware company focusing on building the middle end compiler pipeline for both tensorflow and pytorch frameworks. I mainly worked on MLIR transformations in the middle end but I also designed and implemented frontend python API for compiling and running models natively in python. This includes integrating our compiler pipeline into torch dynamo. I am still actively working in this role since Nov 2021.Prior entering the field of machine learning, I've had 6 years research experience in a range of applied math. I mainly work with various models of the visual cortex and other neural networks, but the mathematical tools involved include linear algebra, functional analysis, dynamical systems, bifurcation theory, stochastic analysis, Bayesian analysis, group theory, and differential geometry.