Software Engineer
Current* Designed and developed a new network forecasting algorithm, aka Tshirt-Forecast-Model, which is a locality based statistical model. Responsible for writing the unit tests, end to end testing, deployment to production and handling production issues thereafter.* Evaluated and compared the performance of different forecasting models. Identified improvements and iteratively improved the quality of Tshirt-Forecast-Model by 50%. Improved the runtime of the model by 10x, and increased its scale from O(10K) to O(100K) network flows.* Desined and developed the AutoCARE system, a system for automated Network Order management that balances network capacity needs, cost and service impact risk for Google's Product Areas(PAs) such as Search, Ads, YouTube, etc. Responsible for writing the unit tests, end to end testing, deployment to production and handling production issues thereafter. Proposed various metrics to evaluate goodness of the product. Enhanced the customer facing dashboard to display metrics and reports as per their needs.* Designed the interface for incorporating Inorganic demands from PAs into the network forecast. Inorganic Demands are the requirements which indicate changes in network traffic patterns due to various reasons such as new service deployments, temporary redeployments or architectural changes. This involved defining - (1) input mechanism, input format and storage mechanism for the demands. (2) types of Inorganic demands which could be supported. (3) time frame within which the demands will be valid.
