Developed a high-resolution, over one million pixels, conditional 3D microstructure generation system using a combined VAE and Diffusion model architecture to generate new microstructures on demand.
Achieved over 400x performance enhancement in characterizing virtual organic semiconductor devices compared to full physics finite element method based Frameworks. Integrated a surrogate model with an adaptive Bayesian approach for high throughput analysis.
Developed a machine learning model focusing on the chemical aspects of virtual Organic Photovoltaic (OPV) device characterization, enabling rapid characterization and morphology optimization.
Assisted in engineering a backend system using Python, FastAPI, MongoDB, and Docker. This system enabled data exchange across universities, managing over 100 billion data points, with features like data filtering, file and JSON metadata integration, and QR code tracking.