Machine Learning and Statistics graduate (UdeM & Mila MSc, McGill BSc) exploring multi-agent reinforcement learning and interactive AI systems. Co-founder of Omi AI, building cooperative Gymnasium and MuJoCo simulations alongside a self-exploring agentic partner that adapts through multimodal dialogue.
Current Work
Co-Founder & ML Researcher — Omi AI | Montréal, QC | 2024–Present
- Developing a general self-exploring agentic partner that adapts and questions user actions through proactive multimodal dialogue and reflective feedback within a multi-agent RL environment.
- Simulating cooperative and adversarial agent-user interactions in Gymnasium and MuJoCo using self-supervised RL with self-generated random environments and structured tasks.
- Building end-to-end training and multi-turn evaluation workflows to analyze agent-user behavior, alignment, model evolution, and failure modes—testing policies against instability and reward hacking under shifting dynamics.
Prediction Market Tool | Open Source | 2026–Present
- Built a Python CLI analysis tool querying live prediction market APIs to estimate calibrated probabilities, evaluating statistical edge and tracking user forecasting accuracy over time.
AI Trainer & Model Evaluator | Freelance | 2025–Present
- Assessing and ranking LLM responses (RLHF) across logical reasoning, Python code correctness, and safety rubrics to improve alignment and quality, documenting stability issues with multi-turn adversarial prompts and Python verification checks.
- Fact-checking model outputs against primary sources to catch subtle hallucinations and constraint violations under structured evaluation guidelines.
Past Work & Projects
Machine Learning Researcher Intern | Humanware, Longueuil, QC | 2022–2023
- Trained and evaluated object detection models (RegionCLIP), improving out-of-distribution generalization in unfamiliar environments.
- Built an Android voice-guidance prototype for visually impaired users with continual-learning data collection.
Multi-Agent RL in PettingZoo | UdeM & Mila, Montréal, QC | 2023
- Compared DQN and PPO in multi-agent PettingZoo Atari environments; analyzed reward shaping, network architectures, and extensions toward self-supervised and offline RL.
Convolution vs. Attention Generalization | UdeM & Mila, Montréal, QC | 2022
- Benchmarked ResNet, ViT, and ConvNeXt using proxy shape-bias metrics on GAN-stylized images to evaluate representation stability across out-of-distribution visual styles.
Hockey Goal Prediction Service | UdeM & Mila, Montréal, QC | 2021
- Deployed an ensemble deep learning prediction API using Flask and Docker with experiment tracking in Comet.ml and interactive Plotly shot heatmaps.
Toxicity & Fairness Evaluation in Text | McGill University, Montréal, QC | 2019
- Evaluated CNN, LSTM, and fine-tuned BERT/GPT models on toxicity classification (ROC-AUC) while optimizing group fairness across identity subgroups.
Software Engineer Intern | Microsoft, Beijing, China | 2015
- Developed and maintained client-facing web components (HTML, CSS, JavaScript) and prototyped a cross-platform intern social app on Android.
Additional Experience
- Summer Camp Counselor | Préville Summer Camp (Summer 2023) — Led daily robotics, science, music, dance, and sports activities for children aged 5–12.
- Orderly | Grace Dart Extended Care Centre (CIUSSS ODIM) (2020–2021) — Supported residents’ daily care routines and group activities in a clinical team.
- Web Dev & IT Support | Centre Saint-Antoine 50+ (2017) & Age UK (2016) — Built an accessible senior website and led 1-on-1 tech coaching across PC, Android, and iOS devices.
Education
Université de Montréal & Mila | Professional MSc in Machine Learning | 2021–2023
McGill University | BSc in Statistics & Computer Science | 2015–2019
Past Present Future
Time Symmetric Mirror
Aligning with Entropy
Exploring with Harmony
Balance Reflective Embrace
Flow Permeating Space