Relapse after autologous stem cell transplant (ASCT) remains a major challenge in multiple myeloma management, underscoring the need for accessible predictors of treatment resistance. This study evaluated whether baseline complete blood count (CBC) and comprehensive metabolic panel (CMP) markers could predict relapse in kappa- and lambda-restricted myeloma patients using random forest modeling. Heatmaps and variable-importance analyses revealed distinct laboratory predictors across subtypes. In kappa-restricted disease, white blood cell count, lymphocyte counts and percentages, red blood cell indices, and liver enzymes (ALT, AST) were the strongest contributors to classification accuracy. In lambda-restricted disease, lymphocyte percentage, mean corpuscular indices (MCV, MCH), red cell distribution width, ALT, and sodium were the top predictors. These findings highlight potential subtype-specific immune and metabolic features associated with relapse risk. However, limited sample sizes—especially in the lambda subgroup—reduce generalizability. Future directions include expanding cohorts, integrating clinical and molecular markers such as cytogenetics or minimal residual disease, and constructing multimodal predictive models. This work demonstrates the feasibility of using routine lab values combined with machine learning to identify patients at higher risk of relapse following ASCT.