cv

Undergraduate physics student at Université Laval working at the intersection of computational astrophysics and machine learning. Currently preparing for graduate studies in computational astrophysics.

Contact Information

Name Alex Baker
Professional Title Undergraduate Physics Student · Computational Astrophysics & Scientific Machine Learning
Email albak1@ulaval.ca
Location Québec, QC
Website https://phd-brown.github.io

Professional Summary

Undergraduate physics student at Université Laval working at the intersection of astrophysics, machine learning, and scientific computing. My work focuses on building interpretable and physically grounded pipelines for astronomical survey data, with current projects in stellar spectroscopy, galaxy morphology, and Bayesian cosmological inference. Preparing for graduate studies in computational astrophysics, with a particular interest in large astronomical surveys and physics-driven machine learning.

Experience

  • 2025 - 2025

    Québec, QC, Canada

    Teaching Assistant — Classical Mechanics (PHY-1005)
    Département de physique, de génie physique et d'optique, Université Laval
    Academic support for undergraduate students in mechanics and special relativity.
    • Explained complex physical concepts and supported students in solving mechanics and mathematical physics problems.
    • Contributed to assessment and correction of undergraduate problem sets.

Education

  • 2023 -

    Québec, QC, Canada

    B.Sc.
    Université Laval
    Physics
    • Focus on astrophysics, scientific computing, and machine learning applications in physics.
    • PHY-3500 — Course research project on dimensionality reduction and interpretability for stellar spectral analysis using LAMOST DR5 × Gaia DR3 data; presented at the PHY-3500 departmental symposium, Musée de la Civilisation de Québec, Winter 2026. Supervised by Prof. Antoine Allard.

Publications

  • 2026
    Manuscript in preparation

    Manuscript in preparation; target venue under consideration.

  • 2026
    PHY-3500 Course Report & Symposium Contribution, Université Laval

    Course report and symposium contribution supervised by Prof. Antoine Allard. Co-authors: Justine Jean and Nérimantas Caillat.

  • 2025
    ξ Dark Energy · BAO MCMC: Exploratory Bayesian Inference for Dark-Energy Models
    Research software release

    Research software release, v0.1.0-exploratory.

Skills

Programming & Scientific Computing: Python, NumPy, SciPy, pandas, Astropy, Matplotlib, Plotly, LaTeX, Bash/Linux
Machine Learning & Data Analysis: scikit-learn, XGBoost, PyTorch, SHAP, UMAP, HDBSCAN, DINOv2, EfficientNet, Monte Carlo Dropout, Weights & Biases
Astronomy & Cosmology: FITS processing, LAMOST DR5, Gaia DR3, SDSS, Pantheon+SH0ES, DESI BAO, emcee, VizieR, CDS XMatch
Tools & Infrastructure: Git, GitHub, GitHub Pages, Docusaurus, Jupyter, VS Code, CUDA, joblib

Languages

French : Native
English : Fluent — scientific writing and communication

Projects

  • AstroSpectro

    Open-source research project for stellar spectral analysis using LAMOST DR5 spectra cross-matched with Gaia DR3.

    • Developed an end-to-end interpretable machine-learning pipeline for stellar spectral analysis using 43,000+ LAMOST DR5 spectra cross-matched with Gaia DR3.
    • Engineered 183 physics-based spectroscopic descriptors, including Balmer-line features, metallic-line indicators, molecular bands, continuum indices, and line-profile measurements.
    • Trained and evaluated XGBoost-based classifiers while explicitly testing the effect of removing instrumental, positional, and potentially leaky metadata.
    • Applied PCA, UMAP, t-SNE, autoencoders, clustering, and SHAP interpretability to study the physical structure of the learned spectral feature space.
    • Current SHAP results suggest that metallicity-sensitive features, including Ca II H and K and Mg b, play a major role in the learned classification structure alongside classical Balmer temperature indicators.
    • Manuscript in preparation; target venue under consideration. Experiment tracking and reproducibility workflows managed with Weights & Biases.
  • AstroVision

    Open-source deep-learning project exploring galaxy morphology, visual representations, and uncertainty-aware classification.

    • Developed a deep-learning pipeline for galaxy morphological classification on Galaxy10 DECaLS, combining DINOv2 visual features, non-parametric morphometrics, and SDSS photometry.
    • Benchmarked baseline CNN, EfficientNet-B0, DINOv2 feature probing, fine-tuning, and late-fusion multimodal classification strategies.
    • Integrated uncertainty-aware methods, including Monte Carlo Dropout and conformal prediction, to identify ambiguous or unreliable morphology classifications.
    • Investigated whether self-supervised visual representations encode physically meaningful galaxy structure, including correlations with morphology and photometric colour.
    • Prepared an internal manuscript draft in A&A style and a reproducible codebase.
  • ξ Dark Energy · BAO MCMC

    Independent computational cosmology project comparing dark-energy parameterisations using Type Ia supernovae and BAO distance measurements.

    • Implemented a reproducible Bayesian MCMC pipeline comparing ΛCDM, CPL, and exploratory oscillating dark-energy parameterisations using Pantheon+SH0ES supernovae and DESI DR2 BAO measurements.
    • Built likelihood functions for supernova and BAO distance observables, posterior diagnostics, corner plots, and AIC/BIC model comparison.
    • Released as v0.1.0-exploratory with explicit documentation of modelling limitations, including fixed sound horizon assumptions, prior-boundary effects, and self-consistency issues in the oscillating model.
  • MOSS System

    Co-inventor of the Modular Orchestrated Storage System, a persistent-memory architecture for LLM workflows.

    • Co-inventor on US Provisional Patent Application No. 63/882,562.
    • Contributed to the design of the orchestration API and developed Python prototypes for persistent-memory workflows.
    • Project currently under technology-transfer and valorisation discussions with Université Laval partners.