Hi! I'm Sam, Montreal based software engineer.
Welcome to my portfolio
Work
2024 - Current
Pytorch
Tensorflow
Linux
Machine Learning Research Intern
2024
Researched the domain gap between real and synthetic data for efficient training of machine learning algorithms.
Following an extensive literature review of state-of-the-art diffusion and adversarial methods, proposed and developed a custom synthetic data generator used to train internal object detection tools, reaching the same level of accuracy as if it were trained on real data.
Designed and implemented a pipeline for grounded image segmentation and unsupervised image captioning to fine-tune generative models.
Produced extensive reports of all results which were presented bi-monthly to the members of the Zebra RnD team.
Following an extensive literature review of state-of-the-art diffusion and adversarial methods, proposed and developed a custom synthetic data generator used to train internal object detection tools, reaching the same level of accuracy as if it were trained on real data.
Designed and implemented a pipeline for grounded image segmentation and unsupervised image captioning to fine-tune generative models.
Produced extensive reports of all results which were presented bi-monthly to the members of the Zebra RnD team.
Research / Implications
Passionate about reverse engineering the brain.
2024 - Current
Undergraduate research student
Supervised by Dr. Christian Desrosiers
Researching and developing novel test time training methods to improve the performance of vision models under domain shift.
Studying the effect of recurrence in vision models for improved efficiency.
Participating to the weekly progress report meetings, where I update the PhD students on my work and participate in the brainstorm sessions.
Researching and developing novel test time training methods to improve the performance of vision models under domain shift.
Studying the effect of recurrence in vision models for improved efficiency.
Participating to the weekly progress report meetings, where I update the PhD students on my work and participate in the brainstorm sessions.
Projects
2024
Keras
ROS
Implemented a deep learning based pose detection algorithm to enable visual command interpretation for Club Capra's autonomous robot.
The underlying model is designed around a fine-tuned YOLOv8 as the feature extractor with a custom classification head.
The robot acheived 98% accuracy and competed in the 2024 edition of European Land Robot Trial in Germany.
The underlying model is designed around a fine-tuned YOLOv8 as the feature extractor with a custom classification head.
The robot acheived 98% accuracy and competed in the 2024 edition of European Land Robot Trial in Germany.
Pose detection
2023
Python
Tensorflow
Keras
Scikit-learn
Implemented a CNN, from scratch, for an age prediction task trained on the FG-NET dataset.
Age prediction
2023
Python
Scikit-learn
Implemented a multilayer perceptron for an emotion prediction task using the FER dataset. Features were extracted using Gabor filters. LDA was then used for dimensionality reduction. Despite the poor results, this was good introduction to the fundamentals of deep learning.
Emotion prediction
2023
Python
Implemented a detective game using a basic linguistic inference model. The computer acted as an agent that was initialized with a few starter facts and was tasked with searching a house for clues on a murder case. As the agent made through the different rooms, they could ask the player questions such as "Where was X suspect an hour before the murder?" via the CLI. After having accumulated enough clues, the agent was able to make a deduction and solve the case.
2023
2022
Play
Work hard, play hard. Here are a few things I like to do to keep my mind fresh.
Music
2013 - Current
Ableton
Via various aliases, I've released music on record labels across the globe, grossing tens of thousand of streams. I've also organized and played multiple concerts.
Item 1 of 8
Cycling
2011 - Current
I'm most likely more agile on two wheels than I am on two feet. I started riding mountain bikes at 14 years old and it didn't take long for me to take an interest in racing. I participated in provincial cups until an injury led me to leave the competitive scene behind. It has since been my way of keeping my physical and mental health in check, in addition to nurturing my connection with nature.
@ Guillaume Troquet Photography