Type 2 diabetes (T2D) is highly heterogeneous, yet the molecular basis of subtypes, particularly normal-weight diabetes (NWD), remains poorly characterized. Extracellular vesicles (EVs) carry signals from multiple metabolically active tissues and protect cargo from degradation, thereby reflecting systemic metabolic heterogeneity. Following ExoTIC-based isolation and comprehensive validation including transmission electron microscopy, nanoparticle tracking analysis, flow cytometry, and Western blotting, we profiled EVs, from Asian normal-weight (A-NWD), Asian overweight (A-OWD), Non-Hispanic White normal-weight (W-NWD), and Non-Hispanic White overweight (W-OWD) T2D patients (n=65) using multimodal surface-enhanced Raman spectroscopy (SERS) and RNA sequencing. SERS revealed spectral fingerprints that distinguished the four subgroups by BMI and race/ethnicity in this cohort. EV-RNA sequencing (n=39) identified differential microRNA (miRNA) expression patterns across subgroups: miR-208a and miR-132 were elevated in A-OWD, while miR-484 was upregulated in A-NWD patients. Notably, A-NWD and W-OWD groups showed partially overlapping EV-associated molecular features in unsupervised analyses, suggesting that population-specific BMI thresholds may mask shared metabolic states in this cohort. This multimodal EV profiling framework provides a structured approach to characterize subtype-associated T2D heterogeneity using EV-associated spectral and miRNA features in a clinically stratified cohort.