Arjun
Dube

Lead the development of our SDK which comprises all aspects of the software development lifecycle from brainstorming ideas to approving designs to writing and reviewing code. Due to the very high performance requirements for streaming multiple 4k 90fps video streams this entails working with highly optimized C++ code and having in-depth knowledge of the video pipeline on the server and client. Often this also requires interop with other languages to unlock the capabilities of a particular client device so our codebase contains Swift, JAVA and Typescript as well as shaders for the various client and server graphics runtimes

Work closely with HMD manufacturers to bring up CloudXR client SDKs for their platforms. This includes collaborating with their engineering teams on optimizing performance and quality, getting APIs we need into their OSs and jointly shipping code. Most recently this resulted in a landmark collaboration between Apple and NVIDIA which was announced at the GTC '24 keynote and entered general availability December '24. Other ongoing partnerships include Meta, HTC and Pico

Drive developer engagement of our SDKs by leading the teams that create documentation, write samples and distribute our components via GitHub. I also guide the creation of in-person labs for example at Siggraph 24' and GTC 25' which drew over 300 developers who have since been running over 1000 CloudXR sessions per week as they develop their own apps and services

Lead the development of our first party integration with NVIDIA Omniverse to enable out of the box XR streaming capabilities without installing a separate driver. This integration involved leading a cross-organizational team to solve the challenges of rendering photorealistic content and compressing it down to a format that could be delivered it over an internet connection. We had to make use of DLSS, RTX and NVENC in novel ways to accomplish our goals of rendering 6k per eye at 45 fps and re-rendering it at 90 fps on the client to produce content that seamlessly blended into the real world

Ideated and lead the development of a patent pending approach to depth based reprojection using a locally rendered mesh to represent remote rendered content. This mesh can be updated less frequently than the device framerate while still yielding world locked, high quality content and can also efficiently defoveate the streamed content saving memory bandwidth and GPU time on the client device. This approach can also be implemented across a variety of client frameworks without underlying access to their system compositors which is very beneficial on more restrictive platforms that assume local rendering

Prior to leading the team I developed several core pieces of our SDK which have been widely adopted by our partners and developers including, encryption and authorization, bi-directional low-latency audio, QoS optimizations and pose prediction

Hired and onboarded a team of eight engineers which involved creating a HackerRank test focused on Spatial Computing and playing the role of a "bar-raiser" in our interview loops to make sure we were assessing the right skills

Created the high level product architecture and validated it with our Technical Fellow community and worked with individual engineers to create their detailed functional specifications to establish a technical roadmap for our funding period (three years)

Participated in customer briefings (Gulfstream, ARAMCO, etc..) and partner discussions (NVIDIA, Facebook, MagicLeap, etc...) to provide technical expertise with the aim of creating and solidifying these relationships

Created a set of contribution processes (Coding standards, reviews and unit/functional testing) and established tooling (Unity, Gitlab etc...) that would allow us to execute in a space totally new to VMware

Owned on time delivery of releases for major milestones such as VMworld 2019 where we presented our augmented workflows solution to the VMworld customer/partner community and Michael Dell

Provided technical leadership to the team including thought leadership (Two filed patents), making high impact technical decisions (When to use and contribute back to OSS for example VRTK), conducting code-reviews for major features and mentoring junior employees (three interns)

Developed a patent pending cloud based object recognition solution that would enable Augmented Reality devices to stream frames from their camera to a server which would run inferences using CV and then reproject those bounding boxes back into three dimensional world space and render on device

Developed a patent pending schema based approach to create both VR immersive training and AR workflow guidance from a single source of truth that establishes the components of the workflow, their spatial relationships in an assembled state and the steps required to connect them

Created a POC demonstrating streaming a Virtual Reality application using the NVIDIA CloudXR protocol from a virtual machine running in VMware's SDDC with a virtual GPU attached. Optimized the performance of the virtualization layer to achieve two simultaneous sessions on a single VM at 60fps each

Created a GoLang microservice which ran open source Terrafrom providers as a scalable cloud service that formed the backbone of our desired state configuration engine for public cloud services. This enabled us to roll out support for a new service in a matter of days instead of weeks

Contributed to and acted as a team-wide expert on our homegrown Java microservices framework Xenon (Now open sourced) that unified load balancing (Custom), asynchronous business logic (Netty) and data storage (Lucene+RAFT) into a single node which could then be scaled horizontally to meet demand

Authored Lucene document schemas that would be used for backend data storage to ensure that data could be queried as quickly as possible while maintaining clear architectural boundaries between components

Created a zero-downtime upgrade system that would allow data migration from blue to green nodes, while intelligently load balancing the traffic based on the "owner" of a document

Utilized 20% time to create a Virtual Reality application to enable visualization and management of the datacenter at scale (10,000+ VMs) which was presented on stage at VMWorld 2017 by our CEO Pat Gelsinger

Created a patent pending plugin based approach to CI/CD artifact management that would allow Code Stream to integrate with multiple repositories like Jfrog Artifactory

Open sourced the SDK for creating similar plugins so that other external partners (Puppet, Chef, ...) could extend CodeStream CI/CD pipelines with their own building blocks

Migrated UI framework from Google Web Toolkit to ExtJS and modernized the look and feel of the vRealize suite of products

Became a member of our University Propel mentoring program which involves working closely with a hand picked new college graduate over an extended six month internship with the goal of placing them in a full time position at VMware

Re-architected a number of legacy Win32 libraries to cross compile on Windows, iOS and Android to enable a cross platform Remote Desktop Client using one codebase

Created a framework to allow drawing and displaying user interface elements on both the iPhone and the iPad using a single codebase

Designed and implemented a suite of UI controls that would allow users to interact with keyboard/mouse driven windows desktop naturally while using a touch driven iOS device

Modified a platform independent HTTPS transport stack to support federated authentication using a proprietary authentication scheme which is slated for use across all our platforms

Developed patented high speed image processing algorithms for the Remote Desktop Windows client using Intel SSE3 vector instructions to enable doubling the compression of our codec with no subjective quality loss and a less than 1% increase in decoding time

Implemented a network based setup for Microsoft Small Business Server comprising a web based front end driving an OS setup bootstrap which applied the settings over a reboot

Architected an HTML5 front end framework on top of Microsoft’s ASP.Net to enable writing an AJAX-enabled, multi-browser, lightweight UI that could be written using Microsoft tools

Wrote a build system on top of Microsoft Team Foundation Service optimized for building, automatically testing and deploying live services currently in use by a team of 30+ developers

Was part of a team wide initiative to come up with best practices and methodologies for agile development in order to enable Windows Server to build services with weekly release cycles alongside in-box products with yearly cycles from a common codebase