DNA-Vault is a novel PyTorch-based neural network architecture designed to solve Catastrophic Forgetting in Continual Learning. Traditional neural networks suffer from weight interference when learning new data. DNA-Vault mitigates this by dynamically partitioning its layers into isolated "vaults". Using a custom Prototype Routing mechanism, the model operates completely task-free during inference. It autonomously infers which task it is solving by calculating the L2 distance between the incoming normalized input and dynamically updated task prototypes, routing the data to the correct vault.