Few relevant & completed projects:

DARWIN: Interpretable Machine Learning for Quantum Chemistry

  • Accelerated search for materials with user specified set of target properties
  • Graph neural network architecture that learns and generalizes from small datasets
  • Automated discovery pipeline for interpretable design rules

Crystal Site Feature Embedding

  • Developed a new representation which achieves state of the art accuracies for bandgaps(0.1 eV) and energies(0.007 eV/atom).
  • Achieves an unprecedented acceleration factor of 10101 for materials and composition exploration
  • Density Functional Theory (DFT) based exploration and training.

Quantum Support Vector Machines

  • Brief summary of how we can lever Quantum Computing to accelerate SVMs
  • Two variants in implementation: based on NISQ devices and large numbered qubit devices

Investigation of generalization in deep neural networks

  • Investigative study of how and when we have learning and when we have simple memorization of the training data
  • Bayesian Evidence as a tool to explain the observations and to predict if we will observe good generalization

Genetic variants classification

  • Exploration of various machine learning strategies to find the best model which performs best in predicting how likely a genetic variant is have conflicting clinical classifications.

NLP inspired model for designing a successful Kickstarter Campaign

  • Combining Natural Language processing (NLP) with stanadard mahine learning techniques, developed a classifier to predict success accuracy of the campaign based on its description, goal targeted, duration of the campaign, etc.
  • word2vec models to articulate most likely to succeed description for the campaign; (multiple contributions)

NLP to analyze podcasts

  • Using NLTK to do some really fun & cool NLP on one of the podcasts I like a lot.
  • If you have some cool idea which I should try on, please drop me a tweet or message!
  • Ongoing hobby project!