Minyue Dai's CPT Project 2019
In the summer of 2019, I interned at Google, MTV as a software engineering intern (SWE). I got an return intern offer from my past EP internship at Google, so this is my second internship in the same office. The internship takes 14 weeks including orientation, and I worked under supervision of my host and co-host. They will be in charge of designing projects, helping me fit in the environment, and reviewing my performance and project outcomes. In EP program interns work in pairs, but in this SWE program I work individually although collaboration is an important part of my job.
Under Google mobile ads, my team focuses on optimizing ads serving strategy of ads on mobile games. My project is to incorporate user data into the current model in order to better improve and test the model. I mainly write C++ code on distributed system for daily data processing and design the new data structure and models for training. I also use SQL for the data validation part. The first sub-project is to integrate Google Play user rating and comments into the system, and the most challenging part is how to translate comments written in hundreds of languages efficiently and accurately. The other sub-project is to design a model to predict user session, for example, how long user x will stay in app x given all related app and user profiles. We decided to build a recurrent model for this task.
This internship fits in what I learnt in Smith College Computer Science program. I used knowledge from Machine Learning and language processing courses to build models and handle comment translation challenge. Also, since we need to push the data to front-end, the database system course also helps me to be familiar with database and SQL programming.