•Designed and deployed an intelligent predictive autoscaling engine which is estimated to reduce the infrastructure costs for the Salesforce IoT Thunder product by $2.5 million this year and by $50+ million per year after the product’s general release
•The predictive autoscaling engine efficiently scales computing resources in the absence of initial data by conservatively learning to map from load to required resources and then predicting the future system load. The engine uses the models’ confidences in order to understand and adapt to the class of usage patterns it is experiencing, leading to a system that is robust to all usage patterns the IoT platform may encounter
•Salesforce is actively pursuing a patent on this scaling engine built by myself and two other interns
