An ensemble does not automatically improve performance
On raw data, the recorded RF + SGD ensemble reached 0.8279 validation macro-F1, just below Random Forest at 0.8293.
Keep a simple baseline and measure what the added method actually contributes.
I build software and explore how machine learning turns data into useful decisions.
Explore the ideas, interactions, and engineering behind three applied AI projects.
Explore how an agent moves through a grid world, picks up a passenger, and learns the value of a good decision.
A closer look at sequence representation, model evaluation, and the importance of keeping data leakage out.
Turn co-occurrence into a useful recommendation. Explore support, confidence, and lift, one basket at a time.
Three editorial observations from saved educational notebook outputs. The original notebooks are not published here.
On raw data, the recorded RF + SGD ensemble reached 0.8279 validation macro-F1, just below Random Forest at 0.8293.
Keep a simple baseline and measure what the added method actually contributes.
Across three recorded learning rates, both implementations changed markedly; at 0.0001 the saved test accuracies were 38.47% and 66.04%.
Inspect training configuration before attributing a gap to architecture or library choice.
Saved outputs include 39 successes in 100 Monte Carlo episodes and 200 in 200 Q-learning episodes, under differing conditions.
Compare methods with shared starts, limits, episode counts, and multiple seeds.
Experience across banking technology and logistics.
A logistics environment connecting CargoWise, SQL, operational indicators, and dashboards.
A software development environment at Banque Nationale, working with a Java and Spring Boot stack alongside Python, Kafka, Jenkins, and Docker.
Detailed responsibilities and exact dates will be added with the final résumé.
34 skills shown
REST API, BDD, CI/CD, Agile, and Scrum appear as working practices, without invented brand marks or proficiency scores.
My focus is software development, with a growing interest in applied AI. I’m interested in the whole path from a problem to a working system: how data is represented, how software is organized, and how a result is evaluated. My experience spans banking technology and logistics, alongside software engineering at Polytechnique Montréal. This portfolio brings those threads together through projects you can explore. Each case study makes room for the decisions, constraints, and limitations behind the work, as well as the next steps needed to make it stronger.
A little more code on GitHubFor software development, applied AI, or a conversation about the work — my inbox is open.