Ads, Adspam Mobile team

Teaching Assitant for Deep Learning (10707) course ￭ TA for Deep Learning (10707) course (https://www.cs.cmu.edu/~rsalakhu/10707/) taught by Prof. Russ Salakhutdinov in Spring 2019. ￭ Designed theoretic and programming assignments to help students understand Attention based language model ￭ Independently mentored 20-30 students’ team projects in areas of Natural Language Processing and Computer vision ￭ Assisted students in groups by answering questions on Piazza and individually by and holding office hours

Built an advertising recommendation system (Python, Scala) ￭ Applied word2wec to obtain text vectors as features from advertisement texts and implemented logistic regression, support vector machine and a neural network with single hidden layer to predict advertisement labels ￭ Performed cross-validation and external validation, achieved a precision up to 83% for 144-class advertisement classification with neural network ￭ Concatenated user labels extracted from user history with the predicted advertisement labels for advertisement the users have clicked for both seed users and industry users ￭ Collaborated with the algorithm engineer to build an advertising recommendation system according to the relevance between seed user labels and industry user labels using Spark and achieved up to 40% reduce on cost per click