Ismael Mendoza

Education

PhD Physics and Scientific Computing Candidate, Department of Physics, University of Michigan, September 2019 – Present
MS Computer Science, Stanford University, September 2018 – June 2019
BS Physics with Honors and Minor in Statistics, Stanford University, September 2014 – June 2018
Honors Thesis: “No escape: light waves in AdS” (https://purl.stanford.edu/vf208qp2190)

Research

Probabilistic Modeling with Machine Learning in Cosmology, University of Michigan, October 2019 – Present
Advisor: Jeffrey Regier (Statistics)

Lead maintainer and developer of BLISS, an open-source Python package designed to measure visually overlapping (blended) galaxies in state-of-the-art astronomical surveys.
Built probabilistic model to measure blended galaxy images using latest algorithms in variational inference and deep generative modeling.
Created pipeline to train, validate, and test machine learning algorithms on real astronomical images.

Leadership in Open Source Software Development, University of Michigan, June 2020 – Present
Advisor: Camille Avestruz (Physics)

Lead maintainer and developer of the BlendingToolKit, a software tool kit for evaluating performance metrics for detection, deblending and measurement algorithms, applied to images of blended galaxies.
Presented software tutorials at collaboration meetings, which recruited a team of contributors.
Led team to extend user interface, incorporate realistic galaxy simulations, and create additional tutorials and documentation.

Computational Cosmology, University of Michigan, September 2019 – Present
Advisor: Camille Avestruz (Physics)

Using dark matter halo catalogs based on N-body simulations to tie together their dynamical history and snapshot properties.
Created pipeline to easily extract dark matter halo present-day properties, merger tree information, and subhalo information for a random subset of haloes at fixed mass.
Designed and executed statistical model to predict present-day dark matter halo properties from its accretion histories.

Observational Cosmology and Data Analysis, Stanford University, June 2015 – April 2021
Advisor: Patricia Burchat (Cosmology)

Developed a statistical framework for weak gravitational lensing that provides a comprehensive analysis of shape measurement noise bias for blended galaxies.
Assessed the impact of blending on cosmic shear estimation for several astronomical surveys.
Publication accepted to the Journal of Cosmology and Astrophysics (JCAP).

Biostatistics, Stanford University, September 2018 – June 2019
Advisor: Julia Palacios (Statistics and Biomedical Data Science)

Implemented efficient algorithms for calculating the likelihood of phylogenetic trees simulated from coalescent models.
Developed Bayesian statistical framework to calculate the probability of correct classification between two different population size histories for large sample sizes and loci.

Convex Optimization, EPFL, June 2018 – September 2018
Advisor: Nisheeth Vishnoi (Theoretical Computer Science)

Participated in Summer@EPFL CS program at the 'Ecole polytechnique f'ed'erale de Lausanne (EPFL).
Designed and executed a project at interface of optimization, cosmology, and Riemannian geometry.
Developed manifold optimization algorithms to measure galaxy shapes from surface brightness profiles.
Used non-convex optimization techniques to mathematically show the high efficiency of my algorithm.

General Relativity and Field Theory Honors Thesis, Stanford University, June 2017 – June 2018
Advisor: Eva Silverstein (Cosmology)

Developed a framework for understanding scattering processes in manifolds by combining insights from quantum scattering theory, differential geometry, and partial differential equations.
Applied framework to successfully resolve paradox of light waves traveling in Anti-de Sitter space.
Simulated complex wave scattering processes using Mathematica.
Presented work as my undergraduate Honors Thesis to the Stanford Physics Undergraduate Committee and at the Stanford Symposium of Undergraduate Research (SURPS).

Publications

Mendoza, I., Mansfield, P., Wang, K., Avestruz, C., “MultiCAM: A multivariable framework for connecting the mass accretion history of haloes with their properties”, Monthly Notices of the Royal Astronomical Society, Volume 523, Issue 4, August 2023, Pages 6386–6400, https://doi.org/10.1093/mnras/stad1768.
Mendoza, I., Liu, R., Hansen, D., Zhao, Z., Pang, Z., Avestruz, C., Regier, J., and LSST Dark Energy Collaboration, “Simulation-Based Inference for Probabilistic Light Source Detection, Deblending, and Measurement”. Submitted to the Dark Energy Science Collaboration (DESC) for internal review.
Wang M.^*, Mendoza I.^*, Wang C., Avestruz C., Regier J., “Statistical Inference for Coadded Astronomical Images”, arXiv:2211.09300. Accepted to the Machine Learning and the Physical Sciences Workshop at the 36th conference on Neural Information Processing Systems (NeurIPS).
Hansen, D.^*, Mendoza, I.^*, Liu, R., Pang, Z., Zhao, Z., Avestruz, C., and Regier, J., “Scalable Bayesian Inference for Detection and Deblending in Astronomical Images”, arXiv:2207.05642. Accepted to the ICML 2022 Workshop on Machine Learning for Astrophysics.
Sanchez, J., Mendoza, I., Kirkby, D. P., Burchat, P. R., and LSST Dark Energy Science Collaboration, “Effects of overlapping sources on cosmic shear estimation: Statistical sensitivity and pixel-noise bias”, Journal of Cosmology and Astroparticle Physics, vol. 2021, no. 7, 2021. https://doi.org/10.1088/1475-7516/2021/07/043.
Mendoza, I., Sainrat T., Boucaud A., Biswas B., Guinot A., Paillasa M., Joseph R., Burchat P., Kamath S., Buchanan J., Avestruz C., “BlendingToolKit: A framework for evaluation of galaxy deblenders”, Monthly Notices of the Royal Astronomical Society, in prep. to be sumbmitted to Journal of Open Source Software (JOSS).

Note: Asterisks denote equal contribution.

Teaching

Course development at the University of Michigan, Ann Arbor, MI
- Course: Physics 104: Introduction to Python Programming, July 2022
- Helped with the development of course materials for this brand new course and midterm/final project.
Instructor, Summer Program in Quantitative Methods for Social Research, University of Michigan, July 2022
- Course: Introduction to Python
- Designed and executed a 10-day bootcamp to introduce Python to newcomers.
- Emphasized a hands-on approach using Jupyter notebooks and encouraged interaction during lectures.
Statistics Teaching Assistant at the University of Michigan, Ann Arbor, MI
- Course: Statistics 507: Data Science and Analytics using Python, August 2020 – December 2020
- Led discussions sections to help students understand Python’s scientific computing stack, relational databases (SQL), and deep learning using pytorch.
- Designed and graded weekly programming assignments.
- Planned and executed a kaggle competition as their final project.
Physics Teaching Assistant at the University of Michigan, Ann Arbor, MI
- Courses:
  - Physics 136: Physics for the Life Sciences Laboratory I, September 2019 – December 2019
  - Physics 141: Elementary Laboratory I, January 2020 – April 2020
  - Physics 453: Quantum Mechanics, January 2022 – April 2022
- Guided students through a series of physics experiments including analysis of their measurements.
- Facilitated group discussions and provided regular feedback on student’s performance.
Physics Teaching Assistant at Stanford University, Stanford, CA
- Courses:
  - Physics 21: Mechanics, Fluids, and Heat, September 2018 – December 2018
  - Physics 70: Foundations of Modern Physics, September 2017 – December 2017
- Designed and graded weekly problem sets, quizzes, and exams.
- Lead weekly problem-solving sessions aimed at reinforcing student’s understanding of lecture.
EPASA: Tutored middle school student in Math and English. September 2016 – June 2018
Habla: Tutored Stanford custodial staff in English 3 hours/week. September 2014 – June 2018

Leadership Roles

Sprint Coordinator for LSST DESC: Organize hackathons and tutorials for the Dark Energy Science Collaboration.
Topical Team Lead for LSST DESC: Organize scientific teams to develop software that accomplishes scientific goals within the collaboration. Currently leading two teams: BlendingToolKit and Bayesian Pipelines (pixels).
Physics Graduate Council at the University of Michigan: Organize social events to help build community at the Physics department.
Life in Graduate School Seminar Council at the University of Michigan: Organize bi-weekly seminars for Physics graduate students to learn about resources at the university that can help them during their PhD. 2022 – 2023

Skills

Python: 7+ years of experience in using Python for coursework and several research projects, including comprehensive knowledge of its scientific computing stack: numpy, scipy, scikit-learn, pandas, matplotlib.
Machine Learning: Extensive experience designing and testing neural networks in pytorch, developing ML pipelines for complex science applications, and knowledge of cutting-edge algorithms including variational autoencoders and normalizing flows.
Other Programming Languages: C/C++, LaTeX, Mathematica, Unix shell, Git, R
Languages: Native Spanish speaker

Honors and Awards

Department Fellowship, Knoller Fund – UofM Physics Department, 2024
Leinweber Center for Theoretical Physics Summer Fellowship – UofM Physics Department, 2023
Walter F. Lewis Candidacy Fellowship – University of Michigan Physics Department, 2022
Science Communication Fellowship – University of Michigan Museum of Natural History, 2022
Computational and Data Science Fellowship – ACM’s Special Interest Group on High Performance Computing (SIGHPC), 2021
Graduate Fellowship – Michigan Institute for Computational Discovery & Engineering, 2021
Enabling Science Award – Large Synoptic Survey Telescope Corporation, 2016 & 2021
Research Grant – Stanford Undergraduate Advising and Research, 2017
Bronze Medalist – 45th International Physics Olympiad, 2014

Presentations

The Blending Problem in Cosmology, Ismael Mendoza, Invited talk at the KIPAC Tea, Kavli Institute for Particle Astrophysics and Cosmology at Stanford University, Stanford, CA, July 2023
Bayesian Light Source Separator (BLISS): Probabilistic detection, deblending and measurement of astronomical light sources, Ismael Mendoza, Invited talk at Statistical Challenges in Modern Astronomy VIII, Pennsylvania State University, State College, PA, June 2023
MultiCAM: A multivariable framework for connecting the mass accretion history of haloes with their properties, Ismael Mendoza, Invited talk at the Baryon Pasting Collaboration Meeting, Yale University, New Haven, CT, May 2023
MultiCAM: A multivariable framework for connecting the mass accretion history of haloes with their properties, Ismael Mendoza, The Co-evolution of the Cosmic Web and Galaxies across Cosmic Time Conference Poster Session, Kavli Institute for Theoretical Physics (KITP), Santa Barbara, CA, February 2023
Statistical Inference for Coadded Astronomical Images, Mallory Wang and Ismael Mendoza, Machine Learning and the Physical Sciences Workshop at the 36th conference on Neural Information Processing Systems (NeurIPS 2022) Poster Session, New Orleans, LA, December 2022
Bayesian Light Source Separator (BLISS), Ismael Mendoza, Dark Energy Science Collaboration (DESC) Summer Meeting Poster Session, Chicago, IL., August 2022
Scalable Bayesian Inference for Detection and Deblending in Astronomical Images, Ismael Mendoza, ICML 2022 Workshop on Machine Learning for Astrophysics Poster Session, Baltimore, MA, July 2022
Machine Learning in Cosmology, Ismael Mendoza, Physics Graduate Student Symposium 2022, Ann Arbor, MI, June 2022
Updates on the Bayesian Light Source Separator (BLISS), Ismael Mendoza, Dark Energy Science Collaboration (DESC) Bayesian Pipelines Topical Team Telecon, April 2022(virtual)
Bayesian Light Source Separator (BLISS), Ismael Mendoza, Dark Energy Science Collaboration (DESC) Winter 2022 Virtual Meeting Poster Session, February 2022 (virtual)
BLISS Update, Ismael Mendoza, DESC Blending Working Group. February 2022 (virtual)
Bayesian Light Source Separator (BLISS), Ismael Mendoza, Dark Energy Science Collaboration (DESC) Summer 2021 Virtual Meeting Poster Session, July 2021 (virtual)
(Updated) Blending ToolKit Tutorial, Ismael Mendoza, Thomas Sainrat, Dark Energy Science Collaboration (DESC) Summer 2021 Virtual Meeting, July 2021 (virtual)
Connecting the Properties of Dark Matter Haloes with Their Growth, Ismael Mendoza, University of Michigan Clusters Group, Ann Arbor, MI, March 2021 (virtual)
Effects of overlapping sources on cosmic shear estimation: Statistical sensitivity and pixel-noise bias, Javier Sanchez & Ismael Mendoza, Collaboration-Wide Presentation for the Dark Energy Science Collaboration (DESC). February 2021 (virtual)
Bayesian Light Source Separator (BLISS), Ismael Mendoza, Dark Energy Science Collaboration (DESC) Summer 2020 Virtual Meeting Poster Session, Chicago, IL. July 2020(virtual)
BlendingToolKit Tutorial, Ismael Mendoza, Dark Energy Science Collaboration (DESC) Summer 2020 Virtual Meeting, Chicago, IL, July 2020 (virtual)
The Blending Problem in Cosmology, Ismael Mendoza, Physics Graduate Student Symposium 2020, Ann Arbor, MI, July 2020 (virtual)
BlendingToolKit: Walkthrough and Future Plans, Ismael Mendoza, DESC Blending Working Group. July 2020 (virtual)

Blog Posts

MathStatBites at SCMA8: Astro Image Processing is BLISS?, Andrew Saydjari for MathStatBites, https://mathstatbites.org/mathstatbites-at-scma8-astro-image-processing-is-bliss/
The Crowded Cosmos: Effects of Blended Galaxies on Cosmic Shear, Katya Gozman for AstroBites, https://astrobites.org/2021/03/20/blended-galaxies-cosmic-shear/