- Replace GPT (current data generation model) with open-source models and run data generation. - Deploy the framework on the Perlmutter machine. - Seed dataset selection.