Description of the image

Piyush R. Maharana

Large Language Models | Machine Learning | Material Discovery

About Me

Project associate at CSIR-National Chemical Laboratory. Working on computational material discovery and retrosynthesis.

Get in Touch

I'd love to connect :)

Education

M.Sc Computational Physics

Central University of Punjab, Punjab, India

November 2021 - June 2023

Courses: Numerical Methods, Electronic Structure Theory, Molecular Dynamics, Density Functional Theory, Atomic and Molecular Spectroscopy

B.Sc Physics

University of Delhi, New Delhi, India

June 2018 - June 2021

Courses: Classical Mechanics, Quantum Mechanics, Electromagnetic Theory, Statistical Mechanics, Nuclear Physics, Mathematical Physics

Skills

Python FORTRAN C++ QUANTICS ORCA VASP LAMMPS Large Language Models Natural Language Processing Neural Network Potentials Mechanistic Interpretability

Experience

Project Associate

CSIR-National Chemical Laboratory, India

Advisor - Dr. Kavita Joshi

August 2023 - Present

Using Natural Language Processing and Large Language Models for scientific data extraction from literature to create databases for training machine learning models for hydrogen storage applications.

Publications

Retrieval Augmented Generation for extraction of data from scientific literature

Preprint, Nov 2024

Abstract: Using large language models paired with retrieval augmented generation, a dataset of metal hydrides for solid-state hydrogen storage was created. The accuracy obtained was 93% in the cases tested. Thus, we demonstrate a pipeline to create datasets from scientific literature at minimal computational cost and high accuracy.

  • Paper
  • HyStor: An Experimental Database of Hydrogen Storage Properties

    Co-author, May 2024

    Abstract: In this work, we introduce the HyStor database, consisting of 1280 metal alloys along with their hydrogen storage capacities (H2 wt%) as a function of absorption temperature. Given the lack of updates in the existing open access HydPark database since 2002, we sourced compositions from recent research articles and various patent documents, resulting in a total of 468 compositions.

  • Paper
  • Quantum Dynamics of H2 formation on interstellar dust grains

    Master's Thesis, Central University of Punjab, India

    Advisor - Dr. Kousik Giri, February - June 2023

    Abstract: The details of H2 formation in the interstellar medium are still unclear. Using MCTDH the dynamics of the formed H2 molecule on a carbonaceous surface was determined to understand the partitioning of the H2 recombination energy in the collinear Eley-Rideal mechanism. Our results show the introduction of the atoms neighbouring the carbon over which H is chemisorbed into the PES of Coronene C24H12, gives the correct energy partitioning with the nascent H2 taking ∼ 85% of the recombination energy.

    Quantum dynamics of coronene
    This image shows the quantum dynamics simulation of H2 formation on a coronene molecule (C24H12). The bright spots represent areas of high probability density for hydrogen atoms, demonstrating the quantum mechanical behavior of the system during the formation process.

    Structure and Kinematics of the Milky Way Galaxy using Atomic Hydrogen

    Conference Paper, June 2021

    Abstract: Presented this paper based on my undergrad thesis in 3rd International Conference on Recent Advances In Fundamental And Applied Sciences (RAFAS), 25-26 June 2021.

  • Paper
  • Rotation Curve of Milky Way
    Rotation curve of the Milky Way Galaxy derived from HI observations. The plot shows orbital velocity versus distance from the galactic center, revealing the presence of dark matter through the flat rotation curve at large radii.

    Spiral Structure and Dynamics of the Milky Way Galaxy

    Undergraduate Thesis, University of Delhi, India

    Advisor - Dr. Vandana Batra, January - June 2021

    Abstract: The hyper-fine transition of atomic hydrogen produces radiation of ν=1.4 GHz which is used to detect intergalactic HI gas in the Milky Way. An observation was carried out from 20° < l < 220° and −20° < b < 20° in steps of 5° using the remotely operated SALSA-Onsala 2.3m Radio Telescope located at Onsala Space Observatory, Sweden. The Perseus, Local and Outer arms of the Milky Way Galaxy were clearly identified. Using the Tangent Point Method the rotation curve till the solar circle was determined giving a value of Vc(Ro) = 240.16 ± 7.03kms−1.

    Face on map of the galaxy
    Face-on view of the Milky Way Galaxy reconstructed from HI observations. The map shows the distribution of neutral hydrogen gas, revealing the spiral arm structure and large-scale features of our galaxy.

    Spectral Index of Synchrotron emission of Crab Nebula

    University of Delhi, India

    Advisor - Dr. Vandana Batra, September 2020

    Abstract: Using archival flux density data ranging from radio to optical from different observations, the spectral index of the nebula in the different electromagnetic regimes was determined.

    Courses and Workshops

    IIIT Hyderabad - 8th Summer School on AI

    July 2024

    Research training event aiming at equipping participants with the relevant advances in the critical and fast developing area of artificial intelligence.

    UCL - Future of Materials Discovery Workshop

    June 2024

    Workshop on AI and machine learning in the physical sciences.

    LLM Hackathon for Applications in Materials and Chemistry

    May 2024

    Teamed up with the University of Liverpool to apply LLMs to improve property prediction for chemical compositions.

    IISER Pune - Natural Language Processing

    January - May 2024

    Attended this semester course on NLP.

    CDAC - Quantum Accelerated Computing

    December 2023

    Attended lectures and hands-on session on Quantum Computing.

    IUCAA - Introductory Summer School in Astronomy and Astrophysics

    May - June 2021

    Attended lectures on various different topics in astrophysics.

    RAD @ Home

    June 2021

    Learned how to make images using the data collected by GMRT and other telescopes and understand UV-optical-IR-radio (RGB-C) images of galaxies to classify galaxies into different subtypes.

    Miscellaneous/Fun

    Neutron Diffusion Equation

    Diffusion equation implosion simulation

    Mayan Hieroglyphics

    Mayan hieroglyphics

    Numerical Methods

    matrix
    Animation demonstrating the Gaussian elimination method for solving systems of linear equations. The process shows how the matrix is transformed into row echelon form through elementary row operations.

    1D Hard Disks

    pin model
    Visualization of a 1D hard disk system, modeling molecular dynamics in a confined space. The pins represent particles that can only move along a line while maintaining minimum separation distances.
  • RDKIT Reaction Visualizer
  • arXiV Paper Explorer
  • RPG world
  • Extracurricular Activities

    Volunteering

    August 2023 - Present

    I teach kids and distribute surplus food to those in need as part of the NGO Robin Hood Army.

    Astronomy Club

    2018 - 2022

    Searching for suspected comets visible in SOHO/LASCO satellite images by participating in the Sungrazer project and taking part in astronomical observations and workshops and lectures on astrophysics and programming at Nehru planetarium.

    International Asteroid Search Campaign

    April - May 2018

    Two provisional asteroid discoveries recorded.

    Supernova Hunters

    2017 - 2019

    Discovered supernovas in Pan-STARRS1 data as part of a citizen science project.

    Flag Counter