About Me
I am a Biotechnology student at VIT Vellore specialized in computational drug discovery and structural bioinformatics. My aim is to bridge the gap between biological intuition and physical reality, replacing empirical 'best-guesses' with computationally efficient and validated models.
I am currently learning quantum-mechanical (QM) methods to validate molecular interactions, transitioning beyond classical approximations to provide thermodynamic ground truth for applications in drug design.
From developing PyTorch libraries with 2,500+ downloads to learning quantum-mechanical validation for complex therapeutic targets, I aim to architect the next generation of computational biology tools.
Pursuing B.Tech at VIT Vellore (2023-2027), maintaining a 9.01/10 CGPA, and dedicated to solving high-impact biological problems through rigorous modeling.
Skills & Expertise
Quantum Chemistry & Design
- Density Functional Theory (ORCA)
- Semi-Empirical QM (xTB & GFN2-xTB)
- Conformer & Tautomer Ensembles (CREST)
- Ligand Strain Energy Validation
- RESP & Quantum Partial Charges
Programming & ML
- Python (NumPy, Pandas)
- scikit-learn, TensorFlow, PyTorch
- R (Bio3D, ggplot2)
- Bash Automation
Structural Bioinformatics
- Ensemble Dynamics (ProDy, Bio3D)
- Validated Docking (Vina, Smina, QVina)
- MD Simulations (GROMACS)
- Free Energy (MMGBSA, gmx_MMPBSA)
- Protein Modeling (AlphaFold, Phyre2)
Genomics & NGS
- RNA-Seq & Single-Cell Analysis
- Variant Calling & Clinical Genomics
- Cancer Somatic Mutation Analysis
- GWAS & DNA Methylation
- BLAST & High-Throughput Pipelines
Drug Discovery
- Virtual Screening Workflows
- ADMET Prediction (SwissADME)
- Cheminformatics (RDKit)
- Thermodynamic Cycle Validation
- Target Discovery (KEGG, COSMIC)
Projects
Ongoing Research Projects
KinaseForge: Generative AI for Kinase Inhibitor Discovery
Manuscript in Submission
Fine-tuned NovoMolGen-157M on kinase-like inhibitors to generate 2.8 million novel kinase-like molecules
Built XGBoost prediction models for 25 kinases to score and rank generated compounds
Created a searchable database enabling multi-target queries — screening compounds active against specified targets and inactive against others
Platform accepts custom SMILES strings to predict activity profiles across all 25 kinases
Launch KinaseForge
Multi-Target Inhibitors for TNBC
Manuscript in Submission
Screening natural inhibitors for TNBC targets with stringent drug-likeness filters
Performing triplicate MD simulations with MMGBSA free energy estimation
Analyzing binding interactions through thermodynamic and dynamics-based benchmarks
Ferroptosis Predictor: Interpretable QSAR Framework
Manuscript in Submission
Developed an interpretable QSAR classification framework distinguishing ferroptosis inducers from inhibitors using a curated, balanced dataset of 1,624 compounds
Engineered 2,092 molecular descriptors and Morgan fingerprints with scaffold-based cross-validation, achieving AUROC 0.9175 and MCC 0.6617
Applied SHAP analysis to identify 75 key predictive features, translating discriminative fingerprint signals into chemically meaningful functional group motifs
Achieved 86.6% alignment when operationalizing discovered patterns as heuristic screening rules on drug-like ChEMBL compounds
View Repository
Inhibitors for Carbonic Anhydrase II
Manuscript in Submission
Developing pipelines for high-throughput screening of Glaucoma therapeutic targets
Validating inhibitory potential through extensive GROMACS MD stability analysis
Natural Plant-Based Diabetes Inhibitors
Manuscript in Submission
Identifying alpha-amylase inhibitors using virtual screening
Utilizing ADMET profiling and MD to evaluate therapeutic safety and efficacy
Characterizing lead compounds for optimized bioavailability and pharmacokinetic profile
Software Development
DynaMune: Protein Dynamics Platform
Integrated platform based on Elastic Network Models and Normal Mode Analysis (NMA)
Implements PRS, domain–hinge detection, and pocket dynamics for structural exploration
Automated extraction of interface stability metrics and non-covalent interactions
View Project
Torchify Python Library
Developed PyTorch utility library for streamlining model training and API management
Published on PyPI with 2,500+ active downloads
Designed to simplify neural network workflows for complex biological datasets
View Project
Automated Virtual Screening
Parallel docking executable for Vina, Smina, and QVina with automated results management
Reduced screening time for high-throughput natural compound libraries
Streamlined structure preparation and scoring for multiple docking engines
View Project
Pose-Rescorer: Deterministic MM/GBSA Rescoring Workflow
Deterministic single-frame MM/GBSA workflow for post-docking ligand ranking using physics-based energy evaluation
Validated across four benchmark protein–ligand systems: EGFR kinase, HIV-1 protease, BRD4 bromodomain, and HSP90
Implements Rapid Perturbation Sampling (RPS) for numerical robustness analysis under coordinate perturbation
View Project