Model deployment/inference, perf optimization, multi-compute software architecture design(handling different types of DNN compute, TRT engines, DLA engines, custom CUDA kernels, CPU logic)

2d and 3d Multi-task deep neural network software design

performance optimization

Light source perception, Landmark detection

Working to ensure platform stability, increase performance across the computer vision stack

Bringing up next generation Autopilot platform

GPU CUDA kernel algorithm inplementation

Brought up and shipped Autopilot HW2 platform

Developed simulations and simulation systems for autopilot