Professional Summary

I am a first-year PhD student at UC Berkeley studying computer science, with a concentration in artificial intelligence. My academic interests lie in efficient machine learning, perception for autonomous driving, and reinforcement learning. Additionally, I have been recognized internationally for social impact and design.

For more, see my resume. Alternatively, see my LinkedIn.

Education

University of California, Berkeley (2018-)
Computer Science (PhD) . Artificial Intelligence

University of California, Berkeley (2014-2018)
Electrical Engineering and Computer Science (BS) . Major GPA: 3.81/4.00
Graduate Combinatorial Algorithms and Data Structures, Graduate Deep Learning, Convex Optimization, Probability Theory and Random Processes, Artificial Intelligence, Machine Learning, Algorithms, Data Structures, Computer Architecture

International Awards

  • International Semifinalist for Adobe Design Achievement Awards ('18) both Social Impact, Commercial
  • Microsoft Imagine Cup World Finals "Big Data" Top 6 Finalists ('18) ~40,000 entrants across 200+ countries, for an indoor positioning application focused on medical devices
  • International Semifinalist for Adobe Design Achievement Awards ('17) both Social Impact, Fine Arts
  • International Top 16 in "Web & Mobile", Design ('17) Semifinalist for "The Rookies Co." design competition with ~9000 entries across 80+ countries, 600+ design schools

Publications

SqueezeDet: Unified, Small, Low Power Fully Convolutional Neural Networks for Real-Time Object Detection for Autonomous Driving
by Bichen Wu, Alvin Wan, Forrest Iandola, Peter H. Jin, Kurt Keutzer
CVPR 2017​ ​.​ ​Winner​ ​of​ ​Best​ ​Paper​ ​Award​ ​at Embedded Vision Workshop . Paper . Code

SqueezeSeg: Convolutional Neural Nets with Recurrent CRF for Real-Time Road-Object Segmentation from 3D LiDAR Point Cloud
by Bichen Wu, Alvin Wan, Xiangyu Yue, Kurt Keutzer
ICRA 2018 . Paper . Video

Shift: A Zero FLOP, Zero Parameter Alternative to Spatial Convolutions
by Bichen Wu, Alvin Wan, Xiangyu Yue, Peter Jin, Sicheng Zhao, Noah Golmant, Amir Gholaminejad, Joseph Gonzalez, Kurt Keutzer
CVPR 2018 . Paper . Code

Experience

Software Engineer . Rex . May 2018 - present

  • building AI scheduling assistant with natural language support for multiple channels--text, email, Slack, Messenger etc.

Machine Learning Research Intern . DeepScale . Summer 2017

  • employed sensor fusion for higher-resolution point cloud prediction with focus on object detection tasks
  • adapted metrics such as Wasserstein distance to quantify performance

Software Engineering Intern . Facebook . Summer 2016

  • worked with selendroid, ADM, and IntelliJ for MyDay (Messenger media-focused redesign), Tincan (encrypted threads) and Picheads (native camera integration)
  • updated internal GraphQL dex, halved number of network requests for photo gallery, centralized Tincan media load

Teaching

Head Student Instructor for Machine Learning
Su'18 . Sp'18 . Fa'17 . Sp'17 . ~500 students per offering

  • lead staff of 30+ uGSIs, readers, academic interns
  • compiled course notes, quizzes, and crib sheets (aaalv.in/cs189)
  • helped reorganize course syllabus and write new content--homework, notes, tests

3-time Head Student Instructor for Discrete Mathematics and Probability Theory
Sp'17 . Fa'16 . Sp'16 . 800+ students per offering

  • 4.83/5.00 rating, 0.3 above department average
  • led staff of 70+ student instructors, readers, and academic interns; academically responsible for 70+ students, 20-hour appointment
  • organized 9 exams and over two dozen course events, across all semesters
  • streamlined course organization, wrote 60-page book, published ~50 quizzes Fa'16, Sp'16

Student Instructor for Structure and Interpretation of Computer Programs
Fa'15 . Sp'15 (tutor) . Su'16 (head tutor) . 1200+ students per offering

  • academically responsible for 60+ students, 20-hour appointment
  • released 30-page compilation, released over a dozen quizzes, wrote and administered 250-person mock midterm
  • core developer for the course's Ok autograder: increased test coverage from 37.1% to 90.7%, enhanced API security, use for 1400+ each semester

Miscellaneous