Sambit Panda profile picture

Sambit Panda

AI Research Scientist @ MATRIX AI Consortium

I’m an AI Research Scientist at the MATRIX AI Consortium where I am developing a end-to-end solution for trauma care using AI. Previously, I completed my PhD at Johns Hopkins, where I was advised by Joshua T. Vogelstein. Here’s a little bit more about me.

Research

Here are some of my favorite research projects. If you want to read more, take a look at the full publication list.

  1. 🗣️ Generative AI for Biomedical Decisions

    A talk giving a basic overview of how to practically get started with generative AI for a biomedical audience. This workshop was done as part of the NIH AIM-AHEAD DICB program.

  2. 📄 Universally Consistent K-Sample Tests via Dependence Measures

    Introduces the idea that the k-sample testing problem and independence testing problem are equivalent up to a transformation of the data.

  3. 📝 Learning Interpretable Characteristic Kernels via Decision Forests

    Demonstrates the kernel derived from random forest is characteristic and develops a hypothesis test based on that fact (KMERF).

  4. 📄 The Chi-Square Test of Distance Correlation

    Derives an approximation to the p-value of distance correlation that bypasses the permutation test with no significant loss of power.

Software

I love solving difficult problems, and often times develop software to help. You can see more in the full software list.

  • iRemedyACT

    An AI-driven tool to inform trauma care policy and an assistant for trauma care physicians at point of care.

    • Python
    • PyTorch
    • scikit-learn
    • LLMs
    • SQL
    • Google Cloud
    • React JS
    • Tailwind CSS
    • Next.js

  • treeple

    Extends scikit-learn decision trees to do oblique splits, manifold learning, hypothesis testing, etc.

    • Python
    • Cython
    • scikit-learn

  • hyppo

    An open-source software package for multivariate hypothesis testing, closing the gap with R.

    • Python
    • scikit-learn
    • CircleCI

  • scipy.stats.multiscale_graphcorr

    Multiscale Graph Correlation is a powerful multivariate correlation test (the first such test in SciPy).

    • Python
    • Cython