Ritesh Kumar - AI for Energy
About Research Publications CV
Ritesh Kumar - AI for Energy

Research

My research combines artificial intelligence, computational chemistry, and experimental validation to accelerate the discovery of materials for sustainable energy applications. I focus on developing data-driven frameworks that leverage machine learning, molecular simulations, and high-throughput experimentation to design next-generation battery electrolytes and catalysts.

Objective: Accelerate the discovery of optimal electrolyte formulations for lithium metal and anode-free batteries using machine learning.

Key Contributions:

  • Developed Electrolytomics: A unified big data approach integrating multiple electrolyte databases (20,000+ entries) for systematic design and discovery (Chem. Mater. 2025)
  • Enhanced Bayesian optimization techniques to handle data-scarce and noisy-label settings in real-world battery discovery
  • Created active learning frameworks that reduced experimental screening by 80% while maintaining discovery accuracy (Nat. Commun. 2025)
  • Developed solvent embedding techniques using graph neural networks to guide electrolyte discovery

Impact: Enabled rapid identification of promising electrolyte candidates, reducing development time from months to weeks.

Objective: Elucidate fundamental mechanisms of how aprotic electrolytes influence electrochemical transformations.

Key Discoveries:

  • Discovered that water clustering in aprotic media modulates activity and enables hydrogenated product formation during CO electroreduction (JACS 2025)
  • Revealed how water hydrogen bonding within nonaqueous environments controls reactivity in electrochemical transformations (Nat. Catal. 2024)
  • Formulated computational descriptors from MD simulations to predict electrolyte performance in CO₂/CO reduction

Techniques: Ab initio molecular dynamics (AIMD), enhanced sampling methods, free energy calculations

Objective: Develop interpretable ML models for accurate prediction of materials properties.

Achievements:

  • Introduced chemical hardness-driven interpretable ML for rapid photocatalyst screening, achieving 90% accuracy with 10× speed improvement (NPJ Comput. Mater. 2021)
  • Created graph neural networks (GNNs) for predicting electrolyte properties including conductivity, viscosity, and electrochemical stability
  • Developed transfer learning approaches to overcome data scarcity in materials discovery

Objective: Establish design principles for efficient catalysts in key industrial reactions.

Major Contributions:

  • Developed electronic structure-based design principles for single-atom catalysts in nitrogen reduction reaction (ChemCatChem 2020)
  • Established universal criteria for Dirac cone splitting in 2D materials for electronic applications (J. Phys. Chem. C 2019)
  • Designed van der Waals heterostructures (C₂N/WS₂) as promising water splitting photocatalysts (J. Catal. 2018)

Collaborations: Extensive experimental validation with groups at IISc Bangalore, IIT, and international collaborators

Objective: Build tools that automate scientific workflows and accelerate discovery.

Software Projects:

  • Automated conversion of STEM images to crystal structures using LLMs and computer vision
  • Bridges the gap between microscopy and atomic-scale modeling
  • GitHub Repository
  • Award: 2025 Visionary Award at LLM Hackathon for Materials Science & Chemistry
  • Autonomous scientific data extraction from literature using LLMs and AI agents
  • Reduces manual data curation time by 95%
  • GitHub Repository

My approach integrates three pillars:

  1. First-Principles Understanding: Using quantum mechanics and molecular simulations to understand fundamental mechanisms
  2. Data-Driven Discovery: Leveraging machine learning and large databases to accelerate materials screening
  3. Experimental Validation: Close collaboration with experimentalists to validate predictions and iterate designs

This synergistic approach has led to 22 publications in high-impact journals and 3 manuscripts under review/preparation.

  • Autonomous laboratories: Developing closed-loop systems that integrate ML, robotics, and simulations for self-driving materials discovery
  • Generative models: Creating novel molecular architectures beyond explored chemical spaces
  • Multi-scale modeling: Connecting atomistic insights to device-level performance
  • Sustainable chemistry: Designing earth-abundant, non-toxic materials for energy applications

Experimental Collaborations:

  • High-entropy alloys for electrocatalysis (with Prof. Krishanu Biswas, IIT Kanpur)
  • CO₂ electroreduction catalysts (with Prof. S. C. Peter, JNCASR)
  • 2D materials synthesis and characterization (with Prof. Pulickel Ajayan, Rice University)
  • Fuel cell catalysts (with Prof. K. K. Nanda, JNCASR)

Computational Collaborations:

  • Battery electrolyte optimization (Prof. Chibueze Amanchukwu, UChicago)
  • Materials databases (Prof. Abhishek Singh, IISc Bangalore)

Funding & Resources

Active Computing Allocations:

  • NSF ACCESS Discovery Allocation (Purdue Anvil: 625,000 CPU hours)
  • University of Chicago Research Computing Center (Midway: 200,000 CPU hours)

Research Grants:

  • AI+Science Research Initiative Fund ($10,000, PI)
  • UChicago Women’s Board Grant ($50,000, Co-I)