Projects

We developed a Python decoding framework aimed at rendering images received from signals transmitted by NOAA satellites. The focus was not only on decoding accuracy but also on optimizing the entire process for increased resilience. Specifically, we addressed challenges such as doppler shifts, sampling rate errors, and sample drops to enhance the overall robustness of the decoding algorithm, ensuring reliable and accurate image extraction from NOAA satellite signals.

I designed a class-based CLI vaccine scheduling app that enables efficient storage and allocation of inventory and timeslots for patients and caregivers. The logic was written in Python and data stored in Microsoft Azure, with the database interface using the pymssql library.

I Implemented and evaluated Variational Autoencoders (VAE) and Generative Adversarial Networks (GAN) with both linear and convolutional architectures on the MNIST dataset. The experiments showed that DCGAN (Deep Convolutional GAN) outperforms other models, generating sharper and less noisy images. Linear models, both in GANs and VAEs, resulted in more noisy outputs, while convolutional layers contributed to smoother results and continuous digit representations, especially benefiting the GAN model.