Home Projects
Projects
Cancel

Projects

Projects

All my projects are organized by research area. Each project describes the scientific problem, approach, methods, results, and impact.

Computational Molecular Engineering

Molecular dynamics simulation, force field development, and understanding the behavior of nanoclusters and complex molecular systems through computational methods.

Cluster-Size Distribution Analysis for Metal Nanoclusters

Tracking aggregation dynamics using graph theory and MDAnalysis

Problem:
How do dispersed metal nanoclusters evolve and aggregate over long MD trajectories? Measuring cluster-size distributions requires identifying connected metal atoms and tracking them frame-by-frame.

Approach:
Developed a complete Python pipeline using MDAnalysis and scipy.sparse to compute distance matrices, identify connected components, and generate temporal evolution plots of cluster sizes.

Methods:

  • Graph connectivity analysis

  • Distance matrix computation

  • Connected components algorithm (scipy.sparse.csgraph)

  • Temporal histogram generation

Results:
Successfully analyzed 90 independent 100 ns MD trajectories of Pd₅₅ clusters in five different solvent environments. Generated publication-quality plots showing aggregation pathways and timescales.

Tools & Technologies:

  • Python

  • MDAnalysis

  • SciPy

  • Pandas

  • Matplotlib

Skills Developed:

  • Trajectory Analysis

  • Graph Algorithms

  • Data Science

  • Scientific Computing

Impact:
Results used for Journal of Chemical Physics manuscript. Provides quantitative metrics for comparing cluster stability across different chemical environments.

Force Field Development for Small Palladium Clusters

Parameterization and validation of Pd₃, Pd₄, Pd₅ force fields

Problem:
Existing force fields for small metal clusters are limited. Need accurate parameters for palladium clusters in aqueous environments to enable long-timescale simulations.

Approach:
Developed custom force field parameters by benchmarking against quantum chemical calculations. Validated against molecular dynamics trajectories and experimental properties.

Methods:

  • QM calculations for reference data

  • LAMMPS force field parameterization

  • MD validation simulations

  • Property comparison (binding energies, solvation)

Results:
Generated validated force fields for Pd₃, Pd₄, and Pd₅ clusters. FF files available in repository. Good agreement with QM results and literature values.

Tools & Technologies:

  • LAMMPS

  • Quantum Chemistry Software

  • Python

  • GROMACS (for comparison)

Skills Developed:

  • Force Field Development

  • Molecular Dynamics

  • Quantum Chemistry

  • Scientific Computing

Impact:
These force fields enable study of larger palladium clusters and provide foundation for reactive MD simulations.

Solvation Effects on Pd Cluster Stability

Water, organic solvents, and mixed aqueous systems

Problem:
Why are small Pd clusters stable in water but less stable in organic solvents? Understanding solvation effects is critical for cluster chemistry.

Approach:
Ran 90+ MD simulations (100 ns each) of Pd clusters in 5 different solvent environments. Analyzed solvation shell structure, coordination numbers, and aggregation propensity.

Methods:

  • MD Simulation (GROMACS, LAMMPS)

  • Radial distribution functions

  • Coordination number analysis

  • Hydrogen bonding analysis

  • Aggregation pathway tracking

Results:
Discovered that water’s hydrogen bonding network stabilizes metal clusters through shell effects. Organic solvents show different coordination patterns.

Tools & Technologies:

  • GROMACS

  • LAMMPS

  • MDAnalysis

  • Python

Skills Developed:

  • Molecular Dynamics

  • Trajectory Analysis

  • Statistical Mechanics

  • Scientific Computing

Impact:
Major finding for understanding cluster chemistry. Impacts rational design of solvents for nanomaterial synthesis and catalysis.

GROMACS Equilibration & Production Pipeline

Automated workflow for MD simulations from build to production

Problem:
Manual GROMACS setup for multiple systems is error-prone and time-consuming. Need a standardized, reproducible workflow.

Approach:
Created bash/Python scripts to automate: system building, energy minimization, NVT equilibration, NPT equilibration, production runs with restart capabilities.

Methods:

  • System parameterization (GROMACS)

  • EM algorithm (steepest descent, conjugate gradient)

  • NVT/NPT equilibration (Berendsen, Nosé-Hoover)

  • Production runs with trajectory analysis

Results:
Complete pipeline from PDB → production trajectory. Applied to 90+ simulations. Reduces setup time by 70% and ensures reproducibility.

Tools & Technologies:

  • GROMACS

  • Bash

  • Python

  • HPC Job Schedulers (SLURM)

Skills Developed:

  • Molecular Dynamics

  • Workflow Automation

  • HPC

  • Scientific Software

Impact:
Enables rapid prototyping of new systems. Facilitates teaching and collaboration.

Custom Water Topology for GROMACS Simulations

OPC, SPC, TIP3P water model parameterization

Problem:
Different water models have different topologies. Need consistent, validated water parameters for solvation studies.

Approach:
Developed and validated GROMACS topology files for multiple water models. Benchmarked against literature values and experimental properties.

Methods:

  • Water model parameterization

  • Topology file generation

  • Property validation (density, heat of vaporization)

  • Equilibration and production runs

Results:
Validated GROMACS topology files for OPC, SPC, TIP3P. Used in all 90+ Pd cluster simulations.

Tools & Technologies:

  • GROMACS

  • Python

  • Topology development

Skills Developed:

  • Molecular Dynamics

  • Force Field Development

  • Scientific Computing

Impact:
Ensures accuracy of solvation environment in all simulations.

Sustainable Process Engineering

Green hydrogen production, biomass conversion, catalysis, and circular chemical manufacturing using process intensification and computational modeling.

Green Hydrogen Production: Pilot Plant Process Engineering

Process modeling and optimization for ONGC hydrogen facility

Problem:
Design and optimize hydrogen production processes for industrial scale. Focus on efficiency, sustainability, and scalability.

Approach:
Developed process models using computational thermodynamics, conducted energy balances, optimized operating parameters for maximum efficiency.

Methods:

  • Process simulation (Aspen Plus)

  • Thermodynamic modeling

  • Energy balance calculations

  • Parameter optimization

Results:
Improved process efficiency and provided recommendations for facility operation. Industrial-scale impact. [Details maintained in confidence]

Tools & Technologies:

  • Aspen Plus

  • COMSOL Multiphysics

  • Python

  • Engineering design tools

Skills Developed:

  • Process Engineering

  • Thermodynamics

  • Process Optimization

  • Industrial Applications

Impact:
Direct industrial application. Improved efficiency of green hydrogen production. Shaped my long-term research vision.

Computational Modeling of H₂ Production Processes

Reactor design and optimization

Problem:
How to design and optimize hydrogen production reactors? What are the key bottlenecks?

Approach:
Modeled hydrogen production using reaction engineering principles. Integrated thermodynamics with kinetics. Optimized reactor design.

Methods:

  • Reaction kinetics

  • Mass and heat transfer

  • Reactor design equations

  • Parameter sensitivity analysis

Results:
Identified key optimization parameters. Provided design recommendations.

Tools & Technologies:

  • Python

  • COMSOL Multiphysics

  • Computational modeling

Skills Developed:

  • Reaction Engineering

  • Process Design

  • Scientific Computing

Impact:

Scientific AI & Data Science

Machine learning applications to chemistry, data-driven discovery, automated data analysis, and scientific computing for chemical engineering problems.

Machine Learning for Molecular Property Prediction

Deep learning models for predicting cluster properties from MD

Problem:
Can machine learning predict molecular properties (binding energy, geometry, reactivity) from MD data? Can this accelerate discovery?

Approach:
Trained neural networks on MD trajectories to predict molecular properties. Tested various architectures (CNN, RNN, Graph Neural Networks).

Methods:

  • Data preprocessing from MD trajectories

  • Feature engineering (structural descriptors)

  • Deep learning architectures

  • Model validation and uncertainty quantification

Results:
Achieved good prediction accuracy. Models can infer properties faster than full MD simulations.

Tools & Technologies:

  • TensorFlow

  • Keras

  • PyTorch

  • Scikit-learn

  • Python

Skills Developed:

  • Machine Learning

  • Deep Learning

  • Data Science

  • Molecular Descriptors

Impact:
Enables rapid property screening and accelerated discovery workflows.

AI Applications in Computational Chemistry

Machine learning pipeline development at Aganitha AI

Problem:
How to systematically apply machine learning to chemistry problems?

Approach:
Developed ML pipelines for molecular property prediction, chemical reaction outcome prediction, and automated data analysis.

Methods:

  • Data pipeline development

  • Feature engineering for chemistry

  • Model training and validation

  • Automation and deployment

Results:
Built production ML systems. Automated complex data workflows.

Tools & Technologies:

  • TensorFlow

  • Scikit-learn

  • Python

  • Data engineering tools

Skills Developed:

  • ML Engineering

  • Data Science

  • Scientific Automation

  • Software Development

Impact:

Scientific Computing

High-performance computing, scientific software development, Python programming, Linux systems, and computational tools for research.

HPC Benchmarking & GPU Acceleration

Performance optimization for MD simulations

Problem:
How to maximize simulation speed on HPC clusters? What’s the best GPU/CPU configuration?

Approach:
Developed benchmark suite. Tested GROMACS and LAMMPS on multiple GPUs and CPU configurations. Profiled code performance.

Methods:

  • Benchmark development

  • Performance profiling

  • GPU acceleration tuning

  • Batch job optimization

Results:
2-5x speedup with GPU acceleration. Established best practices for efficient simulations.

Tools & Technologies:

  • GROMACS

  • LAMMPS

  • CUDA

  • Linux

  • HPC Job Schedulers

Skills Developed:

  • HPC

  • Performance Optimization

  • GPU Computing

  • Scientific Computing

Impact:
Enables 90+ simulations to run in reasonable timescale.

Python Scientific Computing Framework

Reproducible, modular scientific code

Problem:
How to write clean, reproducible scientific code? How to enable collaboration?

Approach:
Developed Python libraries for: trajectory analysis, data processing, visualization. Documented. Version controlled. Tested.

Methods:

  • Software engineering best practices

  • Documentation

  • Unit testing

  • Version control (Git/GitHub)

Results:
Reusable code library. Easy to extend. Enables others to reproduce work.

Tools & Technologies:

  • Python

  • NumPy, Pandas, SciPy

  • Matplotlib, Jupyter

  • Git, GitHub

  • pytest

Skills Developed:

  • Python Programming

  • Software Engineering

  • Scientific Computing

  • Open Source Development

Impact:
Foundation for scalable, reproducible research. Publishable code.

Linux HPC Workflow Optimization

Automated batch job management and data processing

Problem:
Managing 90+ simulations manually is error-prone. How to automate job submission, monitoring, and data collection?

Approach:
Developed bash/Python scripts for: job submission, progress monitoring, automated result collection and organization.

Methods:

  • Bash scripting

  • Job scheduler interaction (SLURM, PBS)

  • Process automation

  • Data pipeline development

Results:
Fully automated workflow from launch to analysis. Processes 90+ results automatically.

Tools & Technologies:

  • Bash

  • Python

  • Linux

  • SLURM/PBS

  • rsync, parallel

Skills Developed:

  • Linux Administration

  • Automation

  • HPC Job Management

  • Workflow Development

Impact:
Reduces manual work. Enables reproducibility. Scales easily.

Project Categories by Research Impact

Machine Learning & Automation

  • ML for molecular property prediction
  • Automated data pipeline development
  • Scientific computing frameworks

Molecular Engineering & Simulation

  • MD simulation methodology
  • Force field development
  • Trajectory analysis at scale

Process Engineering & Sustainability

  • Industrial hydrogen production
  • Process modeling and optimization
  • Computational thermodynamics

High-Performance Computing

  • GPU acceleration
  • HPC workflow optimization
  • Scalable scientific software

Methodological Strengths

Molecular Dynamics Simulation:

  • System building and parameterization
  • Equilibration protocols (NVT, NPT)
  • Production simulations with advanced sampling
  • Trajectory analysis and data extraction

Data Analysis & Scientific Computing:

  • Python-based scientific pipelines
  • Large-scale data processing
  • Statistical analysis
  • Visualization and publication-quality graphics

Machine Learning:

  • Deep learning for molecular data
  • Feature engineering for chemistry
  • Model validation and uncertainty quantification
  • Predictive modeling

HPC & Performance Optimization:

  • Batch job scheduling (SLURM, PBS)
  • GPU computing and CUDA
  • Code profiling and optimization
  • Linux system administration

Code & Resources

Most of my research code is available on GitHub: github.com/alokranjancheme

All code is written with reproducibility in mind:

  • Version controlled (Git)
  • Documented
  • Tested
  • Published alongside findings

Interested in Collaboration?

If you’re interested in any of these projects or related research, please get in touch.

Recently Updated