CV

My current CV is available to download above.

Contact Information

Name Leilany Torres Diaz
Professional Title Research Associate & Data Analyst
Email leilany.torresdiaz@gmail.com
Location Cambridge, MA

Professional Summary

Research assistant and data analyst with experience in neuroimaging pipelines, machine-learning-based neural decoding, and computational psychiatry methods. Experienced in reproducible analysis code, complex fMRI and behavioral datasets, and research support from data collection through presentation.

Experience

  • 2026 - Present

    Cambridge, MA

    Research Assistant
    Phelps Memory Lab, Harvard University and Kredlow Lab, Tufts University
    • Support research activities across the Phelps and Kredlow labs.
    • Complete training in fMRI acquisition and fMRI preprocessing.
  • 2024 - Present

    Cambridge, MA

    Neuro-AI Researcher
    Kempner Institute, Harvard University
    • Built and maintained Python pipelines to preprocess, parcellate, and analyze fMRI datasets.
    • Benchmarked deep vision models against neural responses across visual-cortex regions.
    • Used representational similarity analysis, cross-validated model selection, and bootstrapped confidence intervals in reproducible Jupyter workflows.
    • Managed analysis code with GitHub and documented pipelines for collaborators.
    • Designed behavioral experiments on visual memorability and object recognition, including power analyses and study-design refinement.
  • 2021 - Present

    California and Remote

    STEM and Spanish Tutor
    Independent Tutoring
    • Tutored students in chemistry, calculus, STEM subjects, and Spanish using individualized study plans.
  • 2024 - 2025

    Harvard University

    Organizer
    MLNeuroscience Seminar Series, Kempner Institute
    • Recruited speakers, authored technical newsletters, and coordinated a weekly seminar series for 10 months.
  • 2021 - 2022

    San Jose, CA

    Neurorehabilitation Coordinator
    Norcal Brain Center
    • Collected, organized, and maintained clinical data for more than 20 patients per day across traumatic brain injury, autism, anxiety, and depression presentations.
    • Communicated neurophysiology and neurofeedback concepts to patients and families while supporting protocol adherence and engagement.

Education

  • 2024 - 2026
    Post-Baccalaureate Researcher
    Harvard University, Kempner Institute
    Computational Neuroscience
    • Biological Artificial Intelligence
    • Vision Systems
    • Computational Psychiatry
  • Bachelor of Science
    University of California, Davis
    Neurobiology, Physiology and Behavior

Publications

  • 2025
    Conference on Cognitive Computational Neuroscience

    Torres Diaz, L., and Alvarez, G. A. Poster presented at CCN, Amsterdam, 2025.

  • 2026
    Computational Psychiatry Conference

    Accepted poster presentation.

Skills

Programming and Data Analysis (Advanced): Python, R, NumPy, SciPy, pandas, scikit-learn, PyTorch, TensorFlow, Git, GitHub
Neuroimaging and Signal Processing (Proficient): fMRI, Resting-state fMRI, Functional connectivity, RSA, Nilearn, Brain-Score, EEG, MNE-Python
Statistical Methods (Proficient): SVM, Ridge regression, CNNs, Cross-validation, Permutation testing, Bootstrapping, Bayesian parameter recovery, Hierarchical model fitting
Research Computing (Proficient): Reproducible pipelines, Jupyter, VS Code, Linux, Bash, LaTeX, Slurm

Languages

English : Native speaker
Spanish : Native speaker
Italian : Beginner

Projects

  • rsFC Predictive Modeling Pipeline
    • Built a Python pipeline to classify healthy controls and bipolar-disorder participants using resting-state fMRI connectivity from the UCLA CNP dataset.
    • Used Yeo-Schaefer parcellations, linear SVM classification, and five-fold cross-validation.
    • Achieved 73 percent accuracy with a permutation-test result of p = 0.021 and visualized 77 network-level feature weights.
  • DNN-Brain Alignment Benchmark
    • Evaluated 30 models on fMRI data from 10 participants and 72 stimuli using representational similarity analysis.
    • Used split-half cross-validation for layer selection and bootstrapped confidence intervals.
    • Found AlexNet matched or outperformed modern architectures across ventral-stream regions.
  • Aversive Learning Extension
    • Extended a behavioral paradigm with reinforcement-learning models, Bayesian parameter recovery, hierarchical model fitting, and posterior predictive checks.
    • Introduced and evaluated a hesitation parameter to represent decision uncertainty in aversive learning.