Islamic Art


I'm a recent master's graduate (Data Science) from Worcester Polytechnic Institute with a focus on Mathematical Programming/Optimization. My thesis was on solving nonlinear, non-convex, optimization problems with decision variables exactly for an entire range of constraints at once. I was fortunate to be able to work in a research lab at WPI where I joined a collaboration between WPI, Oxford, and HIAS to better match refugees. I also have experience in software engineering working on performant database engine code at Actian, and a research internship at Ericsson Research in ML/AI working with generative models for point cloud compression.


  • Undergrad Thesis Project

    My undergrad thesis project was to develop two neural networks to classify 4D fMRI data into brain disease categories, evaluate those classifiers, and explain their predictions through tailored visualization techniques.

  • Thesis Project

    My Master’s Thesis Project gave me a taste for research problems in combinatorial optimization and through its completion I learned how to design a solver for the ground up. I also developed some novel enhancements unique to my solution approach.

  • RUTH

    I developed a solution for HIAS, a major nonprofit working in refugee resettlement, called RUTH that has several useful properties.

  • The Multiple Polyak Stepsize Method

    If you are a student of convex optimization, you may be aware of the Polyak stepsize that is optimal for first-order methods. I prove that it can be extended somewhat and show initial results that look promising.

  • A Grammar for SMILES

    Grammar VAE is an important paper in applying deep learning to de-novo drug design. However, extending its approach to more varied datasets requires a new grammar.

  • Most General Mixture Prior

    A lot of people use latent variable models with a location-scale prior. This post explores the most general, in a way “objective”, prior that one might employ when one wants to generalize to a multi-modal distribution.

  • Start of Generative Modeling Series

    I have learned a lot about generative modeling in the process of my research experience at Ericsson Research as well as on a side project wherein I employed it to generate small molecules.