RESUME
Work
experience
Research Assistant, AIST
October 2017 - present
I research low bitwidth CNN model. I implemented C++ framework for CNN which works in any types using "template".
Teaching Assistant, The University of Tokyo
May 2017 - August 2017
I introduced my department to girls students and get in consultation.
I also made a presentation about my career with the guidance of my department.
Software Engineer, HERP Inc.
May 2017 - October 2017
We are developing web tools for HR. I implemented scraping function and some new user interfaces.
STEP Internship, Google Japan
September 2016 - November 2016
I joined the search team. I developed a system using image processing.
Teacher, Keio Tutorial School
September 2013 - September 2016
Teaching math, science(physics, chemistry), English and Japanese for junior high school and high school student
Skills
Image Processing
Linux(Ubuntu / CentOS)
FPGA (Vivado, Xilinx)
Languages
C++ (2 years)
Python (2 years)
Verilog (0.5 years)
Education
M.E, The University of Tokyo
Dept. of Electrical Engineering and Information Systems Graduate School of Engineering
April 2017 - present
I finished some courses about Machine Learning ( http://deeplearning.jp/lectures/aai-2017a/ ), Cloud ( http://cloud.enpit.jp/ ) and Data Science ( http://dss.i.u-tokyo.ac.jp/advance.html ) .
Supervisor: Professor, Tomohiro Kudoh
B.E, The University of Toyo
Dept. of Electrical and Electronic Engineering
April 2013 - March 2017
Thesis title: Create Kakejiku Image Dataset And Examine Visual Feature of Brush Letter
Supervisor: Associate Professor, Koji Yatani
Research
Currently, I'm very interested in machine learning. Right now I'm researching low bitwidth CNN. My greatest curiosity is how many bits are needed for CNN training. I have already implemented CNN's framework in C++ from scratch that can be computed with various representations of numbers. Using that framework, I'm analyzing the distribution of values in CNN to know how many bits are needed.