Abstract: 1. First, we introduce the infrastructure of the experimentation at Uber. (eg. back-end infrastructures, front-end UI, and algorithms on the platform for internal use).
2. We hope to introduce an AI-powered experiment using bandits techniques and show significant results from our EMEA CRM campaign testing a bunch of email subject lines to maximize the customer engagement. We received a significant lift in email open rates and expanded such bandit experiments to entire CRM regions (US/CAN, APAC, LATAM and EMEA).
3. We also worked with Uber Eats data science team to tune the hyperparameters using Bayesian optimization in our continuous experiment framework. Uber Eats ran ML algorithms on the ML platform called Michelangelo and then used the Bayesian optimization to choose the best parameter to maximize conversion rates and revenues. Uber's AI labs is also our collaborator.
Bio: Coming Soon
Senior Data Scientist | Uber
aiforengineers | dataops | west2018talks