Arthur Guillaume
Data Science Portfolio
Data Scientist
Education
-
M.S. Data Analytics Queensland University of Technology (Dec-2024)
-
M.S. Biomedical Sciences Universitée de Montréal (Jun-2020)
Work Experience
Data Analyst Intern @ SAS Institute (Nov-2023 to Dec-2023)
- Assisted with data analysis projects by leveraging SAS tools for cleaning, preprocessing, and visualization. Provided technical support to users and contributed to solving complex data analysis tasks.
- Conducted data cleaning, preprocessing, and visualisation using SAS tools.
- Successfully solved technical issues from users while demonstrating mastery of visual analytics to solve data analysis tasks.
- Won a Kaggle data analysis contest, showcasing advanced analytical and problem-solving skills.
Achievements: Won the SAS® Viya® for Learners Challenge 2023 Technical Stack: SAS Viya, SAS Language
Lead Laboratory Analyst @ Agnico Eagles Mines Limited (Sept-2020 to Oct-2022)
- Oversaw laboratory operations, ensuring accurate data analysis and efficient team coordination. Played a key role in driving data-driven decision-making and maintaining operational continuity during critical periods.
- Designed and maintained dashboards to monitor inventory levels, ensuring efficient resource allocation.
- Conducted extensive testing and analysed datasets to track COVID՞19 infection trends, providing actionable insights to stakeholders.
- Enabled swift decision-making through real-time data visualization and coordinated responses.
- Streamlined workflows, optimized processes, and effectively managed a team to ensure operational continuity and efficiency.
Achievements: I developed dashboards to monitor inventory and case data, maintained consistent working conditions while ensuring compliance with health accreditation standards for the lab, and accurately predicted the required inventory levels for various labs. Technical Stack: Python, Excel, SQL
Research Assistant @ CR-CHUM (Aug-2018 to Jun-2020)
- Utilized advanced statistical analysis and data visualization to uncover insights in glucose homeostasis research, contributing to peer-reviewed publications. Managed and analyzed complex datasets, aligning research objectives with program goals to drive impactful scientific discoveries.
- Analysed and interpreted experimental data using advanced statistical tools, contributing to research on glucose homeostasis.
- Created data visualizations to communicate findings effectively, enhancing the understanding of research outcomes.
- Conducted literature reviews to identify knowledge gaps and align research objectives with program goals.
- Managed timelines and prioritized tasks to ensure the successful completion of experiments and data analysis for a longitudinal study.
- Contributed to two peer-reviewed scientific publications through data analysis and visualization efforts.
- Identified critical research insights by uncovering patterns in complex datasets, validating findings for impactful glucose homeostasis studies.
- Ensured project efficiency and accuracy by optimizing workflows and meeting strict deadlines in a high-pressure research environment.
Achievements: Collected, organized, and analyzed data from various experiments, performed statistical analyses, and created visualizations for scientific publications. Technical Stack: Excel, Prism, SPSS —
Projects
Differential Gene Expression (DGE) Analysis
The project involved a comprehensive comparison of Differential Gene Expression (DGE) tools for RNA-seq analysis, focusing on four major pipelines: edgeR, DESeq2, limma+voom, and NOISeq. Using synthetic RNA-seq datasets, the study evaluated the tools based on sensitivity, specificity, runtime efficiency, and their ability to detect differentially expressed genes under various conditions. By analyzing performance metrics such as concordance, false discovery rates, and computational efficiency, the project provided critical insights into the strengths and trade-offs of each tool, helping inform their application in genomics research and bioinformatics.
Technical Stack: R
Predicting Outcomes and Scores in Rugby World Cup 2023
Developed a comprehensive machine learning analysis pipeline to predict match outcomes and study performance metrics in Rugby World Cup 2023. Leveraged Principal Component Analysis (PCA) to reduce 29 performance indicators into 8 key components, explaining 85% of the variance. Implemented logistic regression for win/loss predictions and regularized linear regression for score difference estimation. Utilized Python libraries for data preprocessing, statistical modeling, and visualization. Demonstrated expertise in sports analytics, feature reduction, and predictive modeling.
Technical Stack: Python, R, statsmodels, general linear model (GLM)
Scalable Web Application to Compute Similarity Scores Between Bacteria
Designed and implemented a scalable web application for analyzing and comparing bacteria by computing similarity scores. This project involved integrating cloud services like AWS for storage (S3) and computation, as well as developing a user-friendly frontend using React and a backend API powered by Node.js.
The application allowed users to upload bacterial data files, perform similarity computations, and view results in real-time, showcasing the ability to handle large datasets efficiently. Although the application is no longer live or actively served, it demonstrates expertise in cloud computing, data analysis workflows, and full-stack development.
Technical Stack: REACT, Javascript, AWS, CSS, SQL, MongoDB
Crowd Counting using Density Maps and DNN
As a Machine Learning Engineer, I developed a cutting-edge crowd counting solution using the Density Map Method. This project leveraged advanced deep learning techniques, including an Encoder-Decoder Neural Network with a pre-trained VGG16 encoder to extract image features and generate accurate density maps. The solution was implemented using TensorFlow and Keras, incorporating Gaussian Kernels for adaptive density map creation and a custom loss function combining Mean Squared Error (MSE) and Structural Similarity Index (SSIM) for precision. The system was trained and evaluated on the Beijing-BRT Dataset, utilizing data augmentation techniques such as random cropping and horizontal flipping to enhance generalization. This work showcases expertise in computer vision, deep learning, and scalable model design.
Technical Stack: Python, Keras
Bacteria Comparison Using k-mer Frequencies
Developed a high-performance application to compare bacterial proteomes based on k-mer frequency analysis. The tool calculates the stochastic probability of 6-mers, 5-mers, and amino acids to determine bacterial similarity.
The project utilized a tech stack that includes C++, OpenMP and High-Performance Computing (HPC). The parallelization was achieved using both explicit and implicit multi-threading with OpenMP, optimizing computational tasks. Additionally, Windows threads were used for task distribution across cores.
The outcome of the project was a significant performance improvement, with a speedup curve generated by testing on various core configurations in HPC environments. This solution is capable of handling large datasets efficiently, speeding up bacterial comparison and facilitating phylogenetic tree construction.
Technical Stack: C++, Bash
Bioinformatics Portfolio
Sequencing Reads Alignment and Genome Annotation Analysis: Aligned sequencing reads to reference genomes using Bowtie2 and analyzed results with featureCounts for genomic insights. Developed Python code to process and visualize genomic data efficiently.
CRISPR Guide RNA Specificity Evaluation: Evaluated CRISPR guide RNA specificity to minimize off-target effects by analyzing potential mismatches using Python and genome datasets (E. coli, S. aureus, etc.). Automated the identification of off-target sites and calculated risk scores based on the Zhang method.
Microbiome Analysis of Smokers and Non-Smokers: Investigated the impact of smoking on microbiome diversity in the nose and throat using Calour (Python library) with BIOM tables, metadata, and taxonomy files. Analyzed microbial community differences between smokers and non-smokers.
Microbial Abundance Classification: Developed machine learning classifiers to predict smoking status based on microbiome data, utilizing scikit-learn in Jupyter/Colab environments. Focused on model interpretability and the application of suitable, simple algorithms for classification based on microbial abundance patterns.
Grazing into the Futur
Developed and implemented a comprehensive time series forecasting model using Vector Autoregression (VAR) and Recurrent Neural Networks (RNNs). Utilized R for data preprocessing, analysis, and visualization, with Python for modeling and algorithm optimization. The project involved applying VAR to capture the linear relationships between multiple time series and leveraging RNNs for learning complex temporal dependencies. Focused on model evaluation, performance tuning, and result interpretation for accurate forecasting in dynamic environments.
Technical Stack: Python, R, Tensorflow
AI-Powered Sokoban Puzzle Solver
Developed an AI-based Sokoban puzzle solver using Python, leveraging search algorithms and state-space exploration techniques. The solver intelligently navigates through complex puzzle configurations by applying AI search strategies such as A* and breadth-first search. The project demonstrates advanced problem-solving capabilities, including the use of heuristic-driven searches to minimize moves and optimize solutions. The integration of AI allowed the solver to handle dynamic puzzle layouts and varying target goals, making it a versatile tool for Sokoban puzzle analysis.
Technical Stack: Python
Automated Fungi Image Classification
The solution for the Automated Fungi Classification project employs a Vision Transformer (DINO) as the feature extractor, utilizing the dinov2_vits14 model from the torch.hub library. The extracted features are passed through a lightweight custom classifier built with PyTorch, featuring a fully connected network with ReLU activation. This approach ensures efficient processing and high accuracy while adhering to the project’s computational and runtime constraints.
Technical Stack: Python, PyTorch, Keras
Object Oriented C# Game Application
Developed a SOS game application in #, utilizing advanced design patterns and SOLID principles. Designed a modular system supporting multiple play modes, including Human vs. Human, Human vs. AI, and AI vs. AI. Integrated Command, Strategy, Composite, Decorator, and Template Method design patterns to ensure scalability and maintainability. Built robust save/load functionality with undo capabilities and dynamic board configurations. Applied Object-Oriented Programming (OOP) best practices to create reusable and extensible components for future game expansions.
Technical Stack: C#