Kehan Wang

EECS 5th Year Master student at UC Berkeley.
Advised by Prof. Avideh Zakhor.

About

Hi! I'm Kehan Wang, a EECS 5th Year Master student at UC Berkeley advised by Prof. Avideh Zakhor. My research interests includes multi-modal deep learning and multiple fields in Computer Vision such as object detection and inpainting. I also completed my undergraduate degree in Computer Science and Data Science at UC Berkeley, and worked with Guanhua Wang at RISELab at the intersection of Machine Learning and Distributed Systems.


I am experienced in both research and software engineering, and hope to become a machine learning engineer in the field of Deep Learning or Computer Vision.

My research experiences challenged me to solve problems creatively, to experiment with potential ideas thoroughly, and to look at other people's work critically.

My past internships focused on software development. They helped me build a solid foundation in software engineering and made me a better coder.

  • Degree: Master of Science, EECS
  • Location: Berkeley, CA 94708
  • Phone: (510)345-7113
  • Email: wang.kehan [at] berkeley [dot] edu

Programming Tools

I mainly use Python in classes and researches, but having used more than 10 different programming languages, I can learn languages very quickly.

Python 100%
C/C++ 80%
Java 80%
PyTorch 100%
TensorFlow 90%
Swift/Kotlin/React-Native 80%

Resume

Education

Master of Science, EECS

2021 - 2022

University of California, Berkeley

Bachelor of Arts, Computer Science & Data Science

2017 - 2021

University of California, Berkeley

GPA 3.91; Dean's List; Honors to date

Honor

1st Place Winner of Cal Hacks 5.0

Nov 2018

Berkeley, CA

  • 1st place in Cal Hacks 5.0, a 36-hour hackathon with ~2000 hackers, ~250 teams from all over the world.
  • Navii is an AR mobile app for indoor navigation.

Leadership

Industrial Relations Chair

Jan 2020 - Dec 2020

Upsilon Pi Epsilon, Nu Chapter

UPE is a Computer Science Honor Society. Industrial Relations committee manages its relations to both the EECS department and our partner companies.

Student Union President

Oct 2015 - Jun 2016

Nanjing Foreign Language School, IC

Organized school-wide student activities. Handles the communication between the student body and the school administration.

Research

Distributed Reinforcement Learning

2019 - 2020

RISELab - UC Berkeley

Distribute a drone’s central controller work onto four individual propellers with Policy Gradient and Imitation learning on MPC using PyTorch.

Wavelet: Deep Learning Speed-up

2020 - 2021

RISELab - UC Berkeley

  • Wavelet: By interleaving different GPU’s peaks and valleys in memory usage, our tick-tock training scheduler can achieve >2x speedup.
  • Paper accepted by MLSys 2021.

Barcode Detection in High Resolution Images

2020 - 2021

BAIR - UC Berkeley

  • Propose a new detection pipeline using a regional proposal network to find regions of most barcodes, fully convolutional network for pixel-wise classification, and OpenCV methods for bounding box extraction.
  • Faster and more accurate than YOLOv4, Mask-RCNN and state-of-the-art barcode detection models.
  • Paper submitted to ICIP 2021.

Multi-modal Semantic Misinformation Detection

2021 - Present

BAIR - UC Berkeley

  • Detect if a Facebook post matches its video content.
  • Construct dataset by random mismatches. Use 3D CNN, BERT to extract video, language features, and embed them onto the same semantic space. Use these embeddings to classify match/mismatch.
  • (Ongoing) Hard mismatches and modal-agnostic.

Work Experience

Research Intern

- Advised by Zihan Zhou, Jia Zheng
May 2021 - Aug 2021

Kujiale(Coohom) - Hangzhou, China

  • Proposed a Transformer-based model to find faces in 2D line drawings by selecting groups of edges.
  • Past papers used search or optimization. Instead, we formulated face identification as a seq2seq problem.
  • Used geometric feature to give our model a stronger prior and achieved > 90% recall and precision.

Software Development Intern

May 2020 - Aug 2020

Microsoft - Redmond, WA

  • Worked on Microsoft Teams, a unified communication and collaboration platform.
  • Developed new features on the Desktop/Webclient: Expiry Status (set “Away” for 30 mins), Set Out of Office auto-reply message.
  • Implemented features using Angular, Typescript, C#.

Mobile Software Development Intern

May 2019 - Aug 2019

Brilliant Home Technology - San Mateo, CA

  • Brilliant Smart Home control panel and mobile app connect all smart home devices together.
  • Developed new features: Alexa Skills compatibility, user preferences persistence, smart shades support, and user home devices administration.
  • Implemented features on both Android and iOS with MVVM architecture.

Full-Stack Software Development Intern

Jun 2018 - May 2019

California PATH - Berkeley, CA

  • Developed PathTransit, a public transit app that supports bus-user location matching for transit services such as BART and AC Transit.
  • Used PostgreSQL, Flask APIs and React-Native

Backend Software Developer Intern

Feb 2018 - May 2018

Simpatica Medicine, Inc - Berkeley, CA

Developed a scalable backend for increasing data needs from machine learning on how different AIDS patients react to different medical treatments.