Kasra Afzali

Software Developer | Data Scientist

Message me on LinkedIn

About Me

I'm passionate about bringing data to solve complex problems. With a unique background in clinical medicine and computer science, I enjoy combining my skills from both fields to bring a fresh perspective to machine learning and software development and create impactful technology.

2024 - 2025
University of Michigan
MS
Data Science
2018 - 2022
University of California Irvine
MD
Doctor of Medicine
2014 - 2017
University of California Davis
BS
Neurobiology
Minor
Technology Management

Experience

2020 - 2022
UC Irvine, Department of Radiology
ML Research Internship
  • Developed CNN models for medical image segmentation enabling automated organ volume calculations, outperforming baseline models detection accuracy from 65% to 99% with 91% specificity.
  • Assisted in data preparation and model evaluation for CNN pipeline processing 10,000+ clinical scans.
  • Aligned clinical goals and technical implementation by translating medical needs into ML specifications.
2018 - 2020
UC Irvine, Department of Ophthalmology
Data Analysis Research Internship
  • Led peer-reviewed study analyzing 20+ years of faculty promotion trends in academic ophthalmology, examining gender and racial disparities in career advancement.
  • Developed data pipeline, data cleaning, and feature engineering of public records across 50+ institutions.
  • Developed multivariate regression and longitudinal time-series models to quantify demographic impacts on academic advancement rates.

Projects

Backgammon Solver

Reinforcement Learning

  • Extended AlphaZero architecture to create novel solution for probabilistic game environments.
  • Designed hybrid architecture with MCTS with RNN priors to efficiently navigate stochastic action space.
  • Optimized self-play training through parallel processing and tree-pruning, achieving 14x improvement in iteration time from 10s to 0.7s.

California Dialysis Regulation

Data Engineering and Statistical Modeling

  • Engineered a multi-stage ETL pipeline in Python to integrate and clean 1M+ semi-structured records.
  • Developed interactive geospatial visualizations and animated time series graphs.
  • Created custom database solution enabling complex cross-table analysis and feature importance modeling.
  • Applied feature importance analysis, to identify key predictors and improve model prediction accuracy.

Diabetes Prediction Model

ML Development

  • Developed end-to-end diabetes prediction ML pipeline, evaluating multiple classifier architectures.
  • Implemented systematic imputation and transformation pipelines and feature engineering.
  • Implemented ensemble models and cross-validation for improved model performance and robustness.

Contact

Message me on LinkedIn