Course Delivery as a Software Problem: Scaling a Course to 1,600 Students

EECS Colloquium

Wednesday, October 19, 2016

306 Soda Hall (HP Auditorium)
4:00 – 5:00 pm

John DeNero

Assistant Teaching Professor
EECS, UC Berkeley

Abstract

Motivated by unprecedented student demand, the UC Berkeley computer science program has tripled its course enrollments in 5 years. The introductory course for majors, CS 61A, is now being delivered to nearly 1,600 students in 47 sections by one faculty instructor and a course staff of 95 undergraduates. This talk will survey the infrastructure and teaching innovations that have allowed the course to scale. Software development has played a critical role, providing new auto-grading strategies and collaboration tools. In addition to a backstage tour of course delivery, this talk will describe methods for evaluating new course components and our plans for offering a fully online version of CS 61A in 2017.

Biography

John DeNero joined the UC Berkeley EECS faculty in 2014. His research in natural language processing focuses on tasks related to statistical machine translation, such as cross-lingual alignment, translation model estimation, translation inference, lexicon acquisition, and unsupervised grammar induction. Prior to his current position, John spent four years as a senior research scientist at Google working primarily on Google Translate, which serves over 1 billion translation requests each day. John has authored over 20 refereed papers in the NAACL, ACL, and EMNLP conferences. He is the primary author of Composing Programs, an online textbook about programming and computer science. He received his Masters in Philosophy from Stanford University and his PhD in EECS from UC Berkeley in 2010.