Working on autonomous missions stack (drone + cloud) and 3D Scan as part of the Inspection and Mapping team

Worked on the 3D scan software: porting it to the new vehicle platforms and developing new features

I work in the Alexa Multitasking team as a software engineer. Our team owns a service which owns and manages skill sessions, powering various multitasking capabilities on Alexa.

My major contribution till date has been in the Skill Resumption program: https://developer.amazon.com/en-US/blogs/alexa/alexa-skills-kit/2020/07/make-it-easier-for-customers-to-pick-up-where-they-left-off-with-skill-resumption. I have been working on handling the onboarding and debug requests for skill developers as part of our private beta program. I am also working on preparing the feature for a public launch: designing services, writing code, testing infrastructure, and dev-ops.

I am a contributor to the Georgia Tech Structure From Motion (GTSFM) [https://github.com/borglab/gtsfm/]. GTSFM is a global structure-from-motion pipeline, which runs can run clusters. We leverage NVIDIA's Dask framework to enable each component of the SLAM pipeline (like feature detection, matching, averaging, bundle adjustment) etc to run on clusters.

I have created the base API of many different modules for GTSFM to create a plug-and-play software where researchers and developers can develop facades for their algorithm and utilize all remaining components of GTSFM to run 3D reconstruction.

I have also ported many different front-end algorithms for feature detection and matching. This includes both deep networks like SuperPoint and SuperGlue, and classical methods like SIFT and Ransac. Utilizing the distributed computing offered by GTSFM, I was able to benchmark 15000+ combinations of front-end (by using different algorithms, networks, and hyperparameters).