Hybrid Transformer based model for EEG Analysis

[parthgoyal974]

Hi @zeydabadi,
I'm Parth Goyal, a second-year Computer Science and Engineering student with a strong passion for data science, deep learning, natural language processing, and system optimizations. While exploring organizations for GSoC 2025, I came across your project and found it particularly engaging, especially given my previous work on a brain wave decoding research project focusing on deconstructing brain waves to figure put what letter the subject was thinking about at the time of data collection.
Currently, I'm involved in a research project focused on genetic subtyping and drug repurposing using graphRAGS. In the past, I've worked on and presented research projects related to building large language models from scratch, quantization, and quantum computing.
In order to gain better hands on understanding of the requirements of this project, I have trained a hybrid model to detect 4 different classes of subject activities. I am now developing an alternative implementation using a Transformer + GRU approach to compare the performance of both architectures and understand the architectural requirements of this project. Once that is done, within the allotted time, I wish to experiment with building a neuroGPT based model but while leveraging brain inspired algorithms like spiking neural networks (SNN) and Hebbian learning. Could you please guide me if my approach on this is correct?
Additionally, could I share the codebase of the model created so far with you on your email for review and discuss the steps forward?

Email: parth.goyal44@gmail.com