• Designed and developed native C++, Python and Android SDK, used for building both in-houseand customer applications on the Industrial IoT Edge, requiring them to be light weight andpolyglot. SDK provides APIs for Configuration, Pub-Sub, Health Alerts.• Bootstrapped the Android version of product as well as built new ML model execution engineand framework with capabilities like packaging and loading any machine learning librariesincluding ones with native components, ability to run models in a sandboxed environment.• Built unified abstraction layer with choices of zmq, sharedmemory transports for the com-munication between customer applications and ML model execution engine on android.• Designed and developed a publisher micro service, which streams data to multiple remotecloud endpoints with features like batching, and processing based on runtime configurablecompute and memory/storage profiles.• Added support for tensorflow lite into our quantization pipeline which tries to optimize mod-els for target platforms.• Scaled 10X our Edge microservices to handle more than 50,000 sensors from 5,000 sensorswhile keeping memory usage same.• Designed a solution for scaling our existing single node base solution to multi node solutionfor High Availability.

• Involved in the design and development of various C++ modules, for facilitating programmatic bidding like building a global cache using kafka and rocksdb for faster response across nodes, ad exchange integrations etc.• Implemented a machine learning pipeline to enable dynamic pricing based on user behaviors and other signals from bid request.• Built a user profile data store abstraction layer which allowed for switching to different key value stores seamlessly.

• Implementing security features Internet Protocol over Peer-to-Peer(IPOP) application backend by enhancing Google’s libjingle library to provide authentication through X.509 certificates.

• Developed support for various HP storage arrays like 3PAR, EVA etc. to provide Disaster Recovery capability to HP's High Availability clustering solution called Serviceguard. Essentially leveraging the storage array management APIs to manage replication, LUN states to allow critical workloads to failover to remote datacenter in case of whole site failure/disaster.• Added support for partner storage arrays, by understanding capabilities provided by their storage management APIs to manage replication and storage states automatically in case of disasters.