Jizhou (June) Wang

An optimistic learner, in harmony with space and time, striving for openness, connection and balance.

I’m a graduate researcher at Mila specializing in Reinforcement Learning.
My current academic interests focus on the application of Multi-agent Reinforcement Learning in fields such as robotics, education and healthcare.

During my free time, I enjoy discovering creative ideas through open-source software and AI such as ChatGPT 🤖, figure skating ⛸️, dancing 💃🕺, playing the piano 🎹 and learning new languages 🌎.

Experiences

Multi-agent Acrobot | Research Project | 2023-Present
- Experimenting with open-source software and frameworks, including Gymnasium, and MuJoCo, simulating cooperative and balancing robotic interactions and movements through self-supervised learning.

Multi-Agent Reinforcement Learning (MARL) | McGill University & Mila | 2023
- Directed a group project focused on the comparative analysis of various Reinforcement Learning (RL) algorithms including Deep Q-Network, and Proximal Policy Optimization (PPO) within the multi-agent Space-Invader Atari game in the PettingZoo environment; presentation shown here.
- Creatively experimented with reward functions, network structure and hyperparameters that led to insights for optimizing the agents’ total return. Identified potential limitations and proposed future research directions including Self-Supervised RL and Offline-RL.

Machine Learning Researcher Intern | Humanware, Longueuil, QC | 2022-2023
- Led the development and fine-tuning of object detection models, leveraging SOTA models such as RegionCLIP to enhance detection robustness, outperforming classical object detectors in out-of-distribution scenarios.
- Implemented an Android demo application with voice guidance for the visually impaired. Proposed novel multimodal models with active data collection for continual learning and adapting to changing data distributions supporting long-term model robustness.

Convolution vs Attention on Image Classification | Université de Montréal | 2022
- Comparatively analyzed deep vision models (ResNet, ViT, ConvNext) based on convolution or attention architectures using a proxy shape-bias metric on out-of-domain stylized samples (generated by a GAN network) to evaluate classification generalization performance; visualizations are shown here.

Hockey Goal Prediction | Université de Montréal | 2021
- Performed exploratory data analysis on NHL play-by-play data by building an interactive 2D shot-heatmap using Plotly and ipywidgets.
- Developed a goal prediction service on Flask with Docker using an ensemble deep learning model based on their ROC-AUC in comet.ml.

Orderly | Grace Dart Extended Care Centre (CIUSSS ODIM), Montréal, QC | 2020-2021
- Provided compassionate care to residents while working in teams managing their personal hygiene and daily routines, while following all safety protocols.
- Improved the mental well-being of residents by engaging in friendly social interactions; organizing groups and partner activities (exercises, social games, puzzles) while actively listening to their needs and inquiries.

Toxicity Detection | McGill University | 2019
- Compared performances across different models such as CNN, LSTM, Transformers and lexical using ROC-AUC metrics for detecting toxicity while maximizing group fairness across different identity subgroups.
- Finetuned transformer models such as BERT, and GPT using word embeddings such as Word2Vec and GloVe.

Irrelevant.ai | ImplementAI Hackathon | 2019
- Designed a movie recommendation system with a two-stage supervised learning model from Fast.ai using collaborative filtering and unsupervised K-means clustering with Scikit-learn.
- Developed a filtering algorithm that lets the user explore beyond the AI recommendation feedback loop while maintaining a positive user rating.

Web Designer | Centre Saint-Antoine 50+, Montréal, QC | 2017
- Designed a minimalistic website utilizing jQuery, Bootstrap and AngularJS, resulting in a responsive interface focusing on enhancing accessibility for elderly clients by optimizing text/content size spacing and color palettes.

IT Support | Age UK, Royal Tunbridge Wells | 2016
- Aided senior clients in a compassionate manner on various IT tasks including web browsing, applications, communication tools and settings, while recommending helpful utilities to improve their autonomy around technology.
-Coordinated appointments and conducted 1-on-1 inquiry-based learning on various operating systems such as Windows, Apple and Android.

Software Engineer Intern | Microsoft, Beijing | 2015
- Maintained web components and websites for new and existing user clients using HTML, CSS, jQuery, Bootstrap, and AngularJS.
- Prototyped “MicroFriends” a social networking app for interns on Android/iOS during the Microsoft Hackathon.

Education

Extras

Youtube Channel 🎹
Figure Skating ⛸️
Brazilian Zouk 💃
A Jazzy Break ☕
Rhythm Game 🎵

Interesting topics in Computer Science:
Multi-agent Reinforcement Learning, Meta-Learning, Continual Learning, Contrastive Reinforcement Learning, AI Safety and Alignment

🩶

Past Present Future
Time Symmetric Mirror
Aligning with Entropy
Exploring with Harmony
Balance Reflective Embrace
Flow Permeating Space