Making computers more efficient has become a desirable challenge, given that recent developments of super large AI models often require large amounts of computing power to properly train and run. As a result, making computers more efficient would require less computing and energy resources to train comparable models, and lowering computing requirements would also make the benefits of model training and inference more accessible to users. For my research project, I'm particularly interested in how we can leverage modifications in both the hardware and software side to increase computational efficiency, and ultimately developing a proof of concept that unites these two concurrent approaches.

‍