Working on CXL technology in the area of memory pooling and switching for data centers. The crucial and a challenging problem that I am working on in a CXL fabric is it’s management and configuration to be done at data center scale.

As one of the three senior engineers in a team of ten, we develop edge IoT gateway, a platform to host IoT applications that control robots such as lathe, paint-guns, welders, smart-grid controllers (RT applications), oil pumps, etc. One of the key customer issues we solve is application consolidation. The gateways are also responsible in collecting time-series data from these robots and perform data analysis and provide real time feedback to the robots to take any corrective actions. Involved in developing solutions for customers’ problems starting from onboarding an application on the controller and getting it deployed on our edge compute. Architect for few key features such as smart-upgrade of cloud controller and edge compute and storage virtualization that provides ability for customers to define custom storage for their applications.

As principal engineer in CTO organization, the goal is to innovate solutions to customer problems and sell the innovation to the business units to be productized. In the process we are required to develop a non-throwaway prototype that can be extended to a scalable product.

CONNECTED CARS: To efficiently and reliably transfer vehicle telemetry to the cloud, architected auto-router that acted as a vehicle’s WAN gateway. Up to 90% compression was achieved on vehicle telemetry. In-vehicle telemetry collection was an RTOS embedded ‘C’ application while cloud collector implemented micro-services architecture using Kubernetes/Docker. Architecture and implementation of the end-to-end solution were my contributions in this project.

IoT (Internet of Things) Onboarding: Eliminated cumbersome process of downloading an app on to a mobile device to enter Wi-Fi credentials and service authorization code for on-boarding IoT device and made it completely seamless. Had an opportunity to play with low-level Bluetooth GATT and ATT interfaces. Modification of the Bluetooth driver to suit our needs and putting together end-to-end solution for the demo to a customer were my contributions.

MACHINE LEARNING:

1. Some mobile clients tend to stick to previously connected APs even in the presence of closer APs. Developed a 4-layer Artificial Neural Network to predict such clients with an accuracy of 87%. Understanding mobile clients behavior helped in designing better handoff algorithms.

2. Improved system security by developing an anomaly detector using power consumption signature of system boot process. Neural network and anomaly detector were developed using python libraries.

CLOUD ORCHESTRATION: Co-architected (with a DE) a service life-cycle management software on Openstack to dynamically scale up and down network virtual functions (NVF), upload day-0 configurations, detect NVF failures in milliseconds running 1000s of NVFs and trigger custom actions. Solution took advantage of Netconf/Yang data modeling for NVF configuration. (http://www.cisco.com/c/en/us/support/cloud-systems-management/elastic-services-controller-esc/tsd-products-support-series-home.html).