Computational Chemist & AI Engineer | QM/MM • Molecular Dynamics • DFT • Machine Learning
I work at the intersection of computational chemistry and AI engineering - applying quantum mechanics/molecular mechanics (QM/MM), molecular dynamics simulations, and machine learning to accelerate drug discovery and understand biological mechanisms at the atomic level.
Currently developing automated pipelines for transition-state location in enzyme catalysis, building hybrid quantum-classical neural networks for medical imaging, and constructing AI-driven industrial optimisation systems that bridge simulation and real-world process control.
From factory floor to closed loop - end-to-end applied ML pipeline processing 261,000+ enriched batch records from a real Bangladesh dyehouse, with an AI Recipe Optimizer (XGBoost/LightGBM) achieving < 10% MAPE on all 5 KPIs and beating the best-quartile human operator benchmark (P25), backed by a 95K-word systematic review of PLC-AI integration architecture.
Keywords: XGBoost LightGBM OPC UA PLC Industrial AI Bangladesh Textile Closed-Loop Control Process Optimisation
Zero-hallucination quantitative economics meets live planetary logistics — multi-agent econometric intelligence platform compiling 507 Damodaran valuation datasets across 18 specialized agents, coupled with sub-meter optical satellite reconnaissance and geospatial maritime AIS radar tracking 45+ crude tankers with direct causal transmission to corporate EBITDA margins and DCF valuations, evaluated against the SOIC 310-Point Scorecard and Stan Weinstein Stage Model, delivered as a self-contained Bloomberg Terminal simulation. Verification: 98.8% audit pass rate (248 PASS / 3 WARN / 0 FAIL), cryptographically sealed with SHA-256 signatures.
Keywords: Multi-Agent Systems Quantitative Finance Damodaran Valuation Macroeconomics Bloomberg Terminal Geospatial AIS Sub-Meter Satellite DCF Simulator SOIC Framework
Automated quantum-classical pipeline that refused to climb the wrong mountain - a hard geometry GATE blocks physically impossible H-transfer poses before any scan launches, then locates the true covalent acylation TS of the COVID-19 drug target at PBE/DZVP accuracy across 4 chained GROMACS + CP2K QM/MM stages.
Keywords: QM/MM CP2K GROMACS SARS-CoV-2 Enzyme Catalysis Transition State DFT Cysteine Protease
Designer solvent, molecular-scale verdict - full CHARMM36/CGenFF atomistic MD of a Menthol:Thymol Deep Eutectic Solvent co-loaded with Ibuprofen and Acetaminophen, revealing that Menthol's OH···O hydrogen bonds preferentially solvate IBU while both drugs diffuse slower in DES than in pure water - consistent with enhanced transdermal retention and reduced systemic absorption.
Keywords: GROMACS CHARMM36 CGenFF Deep Eutectic Solvent Transdermal Drug Delivery RDF Diffusion Molecular Dynamics
🔗 Polyethylene Glycol (PEG-400/600/2000) Polymer MD - Urea Co-solute, Water & Gas-Phase Conformational Analysis
Three molecular weights, two environments, one question: what controls chain collapse? - systematic GROMACS + VMD atomistic MD comparing PEG-400, PEG-600, and PEG-2000 in SPC/E water and gas-phase vacuum with urea co-solute, tracking radius of gyration convergence and frame-wise density to resolve how solvation environment governs the nanocarrier surface coating that controls in vivo protein-corona formation.
Keywords: GROMACS VMD Polyethylene Glycol Polymer Physics Radius of Gyration Nanocarrier Biomaterials Urea Co-solute
Factory-floor statistical science with zero trust in unverified numbers - 28/28 mathematical claims independently re-derived from raw production CSV, finding that shade drives salt dosing with Welch's F = 647.30 (p = 4×10⁻¹⁴³), K-Means blind-rediscovers the factory shade taxonomy at 78.24% accuracy, and H₂O₂ overdose is the single fastest ROI intervention point.
Keywords: Python SciPy scikit-learn Welch's ANOVA PCA K-Means Statistical Verification Industrial Process Optimisation
Every inference, search, email, and voice interaction - entirely on local hardware, zero cloud. Production-grade self-hosted AI assistant with multi-model LLM routing (Ollama, OpenAI, Anthropic), multi-step deep research via SearXNG, RAG with ChromaDB, Whisper STT + Piper TTS voice I/O, CalDAV calendar sync, and sandboxed shell execution - all running locally with JWT-authenticated browser interface.
Keywords: Python Flask WebSockets Ollama RAG ChromaDB Whisper Self-Hosted Privacy-First LLM
From target to binding free energy - CADD workflow collection covering SARS-CoV-2 Mpro covalent docking (Glide IFD + 100 ns MD), CYP3A4 ligand coordination (200 ns Desmond, Fe-N distance 2.12 ± 0.08 Å), PEG-400 multi-RDF solvation analysis, and adaptive biasing force (ABF) free energy calculation (ΔG = −8.4 ± 0.6 kcal/mol).
Keywords: Schrödinger GROMACS NAMD CADD Molecular Docking MD Simulation Free Energy Drug Discovery
Classical backbone, quantum discriminator - ResNet50 transfer learning feeds 6 features into a 6-qubit PennyLane variational quantum circuit (6 entanglement layers), achieving 82-98% accuracy on 6-class oral disease classification and demonstrating measurable quantum advantage over a classical Dense equivalent on this low-dimensional feature space.
Keywords: PennyLane TensorFlow Quantum Machine Learning ResNet50 Transfer Learning Medical Imaging VQC Oral Disease
Engine-agnostic precision at scale - A robust collection of advanced molecular dynamics computational protocols and HPC simulation workflows converted to universal Python configurations. Features hierarchical multiple-timestep integration (up to 5 fs with SHAKE/LINCS equivalents), Particle-Mesh Ewald electrostatics, robust NVT/NPT thermodynamic ensembles (isotropic and anisotropic barostats), and systematic MM/PBSA binding free energy evaluation pipelines.
Keywords: Molecular Dynamics Python GROMACS LAMMPS MM/PBSA Free Energy Simulation Analytics Structural Bioinformatics
| Domain | Tools & Technologies |
|---|---|
| Quantum Chemistry / QM-MM | CP2K, Gaussian, ORCA, PySCF, DFT (PBE, B3LYP), DZVP/cc-pVTZ |
| Molecular Dynamics | GROMACS, NAMD, Desmond (Schrödinger), CHARMM36, GAFF2, OPLS3e |
| Drug Discovery / CADD | Schrödinger Glide (IFD, SP/XP), MM-GBSA, ABF, VMD, PyMOL |
| Machine Learning | XGBoost, LightGBM, scikit-learn, TensorFlow/Keras, PyTorch |
| Quantum ML | PennyLane, VQC, qubit-based hybrid architectures |
| Data Science | pandas, NumPy, SciPy, Statsmodels, Plotly, Matplotlib |
| AI Engineering | Flask, WebSockets, RAG (ChromaDB), LLM routing (Ollama/OpenAI) |
| Languages | Python, Bash, TCL (VMD), PowerShell |
Open to research collaborations, industrial AI consulting, and postdoctoral opportunities in computational chemistry and AI-driven process optimisation.