Machine Learning Projects

Celebrity Cruises RAG Assistant

An intelligent conversational assistant for Celebrity Cruises information, powered by Retrieval-Augmented Generation (RAG). This system combines SQL database queries with vector store retrieval to answer questions about cruise availability, pricing, schedules, and general cruise information using LangGraph and LangChain.

Metadata curation to evaluate the use of LLMs in the context of a systematic review

This project is about collecting metadata from articles in systematic reviews to create clean datasets. The developped program performs automatic data cleaning, definition, and alignment for all these articles. The datasets are used by a Mila - Quebec Artificial Intelligence Institute affiliated M.Sc. student to evaluate the use of LLMs in the screening phase of systematic review.

Data Augmentation for Cell Segmentation

Developed smart cropping and GAN-based synthetic data generation strategies, implemented modality-aware augmentations to address multi-modality cell image variability, and integrated these methods into a U-Net architecture for weakly supervised cell segmentation. Built a complete training pipeline with an initial 21-epoch experimental validation, establishing a foundation for extended training and evaluation of segmentation improvements.

Machine Learning for the Best NHL Pool

The project involved extracting and preprocessing NHL player data using their API, predicting current season points with machine learning models, and building the optimal team using an optimization solver.

Benchmarking NLP Model Performances

This project evaluates and compares various text classification systems capable of identifying paraphrases between sentences, testing the robustness of the models on the QQP and MRPC datasets from the GLUE corpus. Furthermore, the project explores whether biased training can lead to biased performance and experiments with different encoders to identify the most performant and generalizable ones.

Kaggle Competition on Text Classification

Not knowing the context and only having a Bag-of-Words matrix, we needed to predict the binary labels of new documents. We developped a small architecture to organize our submissions and our best model was an Ensemble Learning Classifier using a Naïve Bayes Classifier, an XGBoost Classifier and a SVM Classifier.

Miscellaneous Projects

Automatic Detection of Incoming Cars in a Parking Lot

Physical Arduino system with an ultrasonic sensor that sends mobile notifications using Telegram when a car passes by. The operational system will be deployed next summer at the Aux Volets Noirs vineyard.

FormAPI - Contact Form Email Service

A secure Flask-based REST API that handles contact form submissions via Gmail API. Features OAuth2 authentication with dual-mode support, Docker deployment, and intelligent credential management for seamless local development and production deployments.

Custom Portfolio Website

A modern, bilingual personal portfolio website built with Astro 4.x, TailwindCSS, and DaisyUI. Features include a blog system with content collections, dynamic projects showcase, interactive CV timeline, multilingual support (EN/FR), contact form with custom API integration, and fully responsive design optimized for all devices.

Design and Development of a Website in HTML5

Redesigning the Aux Volets Noirs vineyard website using HTML5, CSS, and Bootstrap to gain experience on these technologies while working on giving the best user experience.

Development of a Mobile Game with the Godot Game Engine

Developing a mobile game using the Godot Engine named PewPewPew: a top-down shooter where players survive against hordes of zombies.

Development of a Discord Bot in JavaScript

Developing a Discord bot in JavaScript for Among Us, automating player muting and channel switching to let dead players communicate while listening to the alive ones.

Development of a Personal Game in Python

Developing a personal game in Python using the Pygame library: a 2D platformer where players gain additional powers each level.

View My CV