Research Assistant on SMART DYEING, an AI-driven closed-loop process control system for knit fabric dyeing, funded under a BIRDI grant (Skills for Industry Competitiveness and Innovation Program, Ministry of Finance, financed by the Asian Development Bank and the Government of Bangladesh).
I build machine learning that runs against physical processes rather than against benchmarks. Textile engineering first (dyes and chemicals), then computer science, which means I work on industrial problems from inside the domain rather than from a dataset someone else exported.
Current focus: physics-informed neural networks for adsorption and process modelling, computer vision for industrial defect detection, and resource-aware model deployment on edge hardware.
| Repository | What it is |
|---|---|
| WCF_Biofilm_Adsorption | Molecular-dynamics-informed and physics-informed modelling of reactive dye adsorption on a waste-cotton PVA/TiO2/cellulose biofilm. Accompanies a manuscript under review at J. Environmental Chemical Engineering. Archived at Zenodo. |
| Catkin_Biofilm_Adsorption | A unified predictive surface coupling Freundlich isotherm with pseudo-second-order kinetics through a mass balance, benchmarked under leave-one-out cross-validation against GPR, random forest, XGBoost and a PINN. |
| PINN | Physics-informed neural network implementations under varying physical constraints and boundary conditions. |
| Catkin_Characterization | ML and ANN characterisation models for the Catkin (Saccharum spontaneum) biofilm. |
- MSc, Computer Science and Engineering, United International University
- BSc, Textile Engineering (Dyes and Chemical), Bangladesh University of Textiles