Bo Yang


University of California, Berkeley (Ph.D.)

2015 - 2019 summer
Mechanical Engineering/Biomechanics
Berkeley Biomechanics Lab


The J. K. Zee Fellowship (2015)
SB3C Ph.D. paper competition finalist (2017)
Signatures Innovation Fellows (2017-2018)
ASME-BED - PhD Level Student Paper Competition and Student Travel Bursary (2016, 2017)

University of California, Berkeley (M.S.)

2013 - 2015
Mechanical Engineering/Ocean Engineering

Dalian University of Technology (B.S.)

2009 - 2013
Naval Architecture and Ocean Engineering
Daoda Enterprise Scholarship (2011)
The Lloyd's Register Educational Trust Scholarship (2013)
General Design of a 10000DWT Coastal Product Oil Carrier
Shape, Arrangement, Cross-Section, Propeller

Work Experience and Projects

Internship at TianMing Data Science Technology
(2017 Winter)

  • Developed deep convolutional neural network (CNN), including resnet and densenet, to conduct pneumoconiosis diagnosis using chest x-ray images.
  • Developed Python classes and scripts for medical image analysis.
  • Developed algorithms for data augmentation and checking label quality.
  • Developed deep CNN models using mxnet and gluon and deployed them for inference (80%+ accuracy).

Graduate Student Instructor (GSI) 5 times at UC Berkeley

  • Including one semester Lead GSI of a 260-student MATLAB programming class (Engineering 7).
  • Lead 6 other GSIs to prepare and grade homeworks and exams, give lab sessions, and hold office hours.

Implementation of a Database in Java
Spring (2018)

  • Implemented persistent B+ trees that mapped keys to records.
  • Implemented table, page, and record iterators.
  • Implemented join algorithms over tables including Page/Block Nested Loop Join, and Sort Merge Join.
  • Cost estimation, maintenance of statistics, and query optimization using System R dynamic programming.
  • Developed a Lock Manager to implement table and page-level locking.

Application of CNN to Medical Image Detection and Segmentation

  • Obtained 1082 slices of 256×256-resolution MRI scans of bovine intervertebral disc (Bruker 7T).
  • Preprocessed images: rotated and transposed images to enlarge total data amount, and labeled each pixel.
  • Improved image classification and segmentation task accuracy drastically to 97%.
  • See more

Parallelized transformation from 3D Matrix to Voxel Mesh

  • Developed an algorithm in C++ that was O(n) complexity.
  • Paralleled C++ code using Cuda GPU.
  • Paralleled C++ code using MPI IO (6 × speed up).

Selected Coursework

Course Title Grade
CS 61B Data Structures A+
CS 286A Introduction to Database System A
CS 267 Applications of Parallel Computers A
CS 289A Introduction to Machine Learning Pass
CS 294-131 Deep learning Special Topics Pass
EE 128 Feedback Control System A+
MEC 232 Advanced Control System A
Math 228A&B Numerical Solution of Differential Equations A+
Math 273 Topics in Numerical Analysis A
ME241 Marine Hydrodynamics A+