Supervisor in discussion for my Master Thesis upcoming winter semester : Björn Helm [bjoern.helm@tu-dresden.de ( Technische Universität Dresden )]
A comparative study migrating a traditional hydrological water-balance model (R/Shiny) into a high-performance Convolutional Neural Network (CNN) framework using PyTorch.
Ongoing projects/researches to align with my Master Thesis this coming winter : Do we always need a GPU? Rethinking Necessity in Modern AI Workloads to Compensate the Water Usage, which will help in climate change. Currently in discussion with my supervisor in TU Dresden : Björn Helm
The original model was developed in R to provide a real-time dashboard for tank dimensioning. I have pasted a few pictures and snippets to under 01_R_Prototype/images to display the UI.
- 01_R_Prototype: Original R-Shiny implementation of the green roof storage tank dimensioning tool.
- 02_PyTorch_Migration: Active development of the tensor-based model, utilizing Autograd for mass-balance constraints.
- 03_Technical_Manual: A Jupyter based guide documenting the mathematical transition from R to Python.
- Domain: Urban Hydrology & Green Infrastructure in Germany.
- Tech Stack: R (Shiny), Python (PyTorch), Google Colab, Jupyter Notebook.
- Scientific Method: Implementing Hargreaves PET equations as differentiable layers.
Status: Phase 2 (Migration) in active development.