Sungsu Lim

I am currently working as a data scientist at Tonal .

I have research experience in different areas of machine learning, including reinforcement learning and computer vision. I am interested in solving everyday challenges to make products and services more efficient. I believe technical skills to apply abstract knowledge to tackle real problems are becoming more crucial.

My life so far has been an interesting journey. Born and brought up in Korea, I was fortunate to have the opportunity to study in various countries including India, Hong Kong, U.S., and Canada. I have just entered the 2nd quarter of my life and I am eager to explore and discover what is out there in this world.

'In their hearts humans plan their course, but the Lord establishes their steps.' - Proverbs 16:9

Contact : ss.sungsulim_at_gmail_dot_com

[Last updated: Apr 25, 2021]


Data Scientist

Tonal [link]
  • Improving workout experience through better weight suggestions and form feedback.
10.2020 - Present
Toronto, Canada

Research Assistant

University of Alberta
  • Research focus on control methods in reinforcement learning.
  • Developed a novel framework for using Q-learning in continuous action spaces.
05.2018 - 09.2020
Edmonton, Canada

Associate Researcher, Reinforcement Learning

Huawei (Noah's Ark Lab) [link]
  • Worked on an applied project to increase automation efficiency using reinforcement learning (subject to NDA).
  • Developed deep continuous control models using PyTorch and Python3.
05.2019 - 08.2019 (3 mo.)
Edmonton, Canada

Software Engineer Intern, Machine Learning

Hive [link]
  • Developed face and logo detection/recognition models for providing comprehensive media analytics on celebrity/brand exposure across TV channels.
  • Worked on customizing deep learning models (Single-Shot Detector, faster R-CNN, R-FCN) and improving data preprocessing procedure using TensorFlow and Caffe.
05.2017 - 08.2017 (3 mo.)
San Francisco, U.S.

Research Intern, Computer Vision

SenseTime [link]
  • Worked for the research team at a unicorn startup developing core computer vision technologies for various clients around the world.
  • Worked on fast depth perception, computing 3D depth and its confidence map simultaneously from stereo images by extending traditional BRIEF feature descriptor to multi-scale.
06.2016 - 12.2016 (7 mo.)
Shatin, Hong Kong

Teaching Assistant

University of Alberta
  • CMPUT 174: Introduction to Foundations of Computing I (Fall 2017, Winter 2018)
  • CMPUT 397: Reinforcement Learning (Fall 2019)
09.2017 - 04.2018
09.2019 - 12.2019
Edmonton, Canada


University of Alberta

PhD in Computing Science (Reinforcement Learning)
Advisor : Prof. Martha White and Prof. Adam White
Voluntary withdrawal to work more closely in the industry

09.2019 - 05.2020
Edmonton, Canada

University of Alberta

MSc (Research) in Computing Science (Reinforcement Learning)
Advisor : Prof. Martha White
Thesis : Actor-Expert: A Framework for using Q-learning in Continuous Action Spaces [paper]

09.2017 - 08.2019
Edmonton, Canada

The Hong Kong University of Science and Technology

BEng in Computer Science
Final Thesis Advisor : Prof. James Kwok
  • Graduated with First Class Honors
  • Study leave for mandatory military service in Korea (05.2013 - 02.2015, 21 mo.)
  • School of Engineering Student Ambassador (2012-2013)
  • Overseas Exchange Program (Northwestern University, Fall 2012)
09.2011 - 12.2016
Kowloon, Hong Kong

American Embassy School

IB Diploma
09.2009 - 05.2011
New Delhi, India

Publications and Preprints

  • Chan A., Silva H., Lim S. , Kozuno T., Mahmood A. R., White M., Greedification Operators for Policy Optimization: Investigating Forward and Reverse KL Divergences . Preprint, 2021
  • Satsangi Y., Lim S. , Whiteson S., Oliehoek F., White M., Maximizing Information Gain in Partially Observable Environments via Prediction Rewards . Accepted to AAMAS, 2020
  • Yasui N., Lim S. , Linke C., White A., White M., An Empirical and Conceptual Categorization of Value-based Exploration Methods . Accepted to Exploration in Reinforcement Learning Workshop. ICML, 2019
  • Lim S. , Joseph A., Le L., Pan Y., White M., Actor-Expert: A Framework for using Action-Value Methods in Continuous Action Spaces . Accepted to Deep Reinforcement Learning Workshop. NeurIPS, 2018
    [paper] [poster]
  • Lim S. , Tai Y., Ren J., Yan S., Liang Y., BRIEF Stereo: Fast Stereo Matching using Multiscale Binary Feature . Work done at SenseTime, 2016


RLControl: Collection of Reinforcement Learning Algorithms in Continuous Action Space


  • This repository contains various reinforcement learning algorithms, including value-based and policy gradient methods for continuous action spaces. This code base was also used for my Masters Thesis.

ArmStrong: Smart Coaching Mat with Real-time Feedback


  • Interdisciplinary Global Product Development project in collaboration with students from different universities in Hong Kong, Korea, and China.
  • The coaching mat keeps track of the number of push-ups and also gives feedback if the forces on the hands are uneven. The feedback is given through a mobile app connected via bluetooth.
  • Used Arduino with ultrasonic sensors, pressure sensors, and bluetooth module.

SmartWalk: Fall-detection Shoe for the Elderly


  • Summer Design Thinking project in collaboration with China Academy of Arts.
  • Based on the theme of 'the elderly', we developed a prototype shoe that detects if the wearer falls down and alerts family members.
  • Used Arduino with accelerometer and gyroscope.