Background/Objectives: Late diagnosis hampers effective treatment of non-small cell lung cancer (NSCLC). This study evaluated whether circulating microRNAs (miRs), miR-155 and miR-3196, measured in liquid biopsy peripheral blood mononuclear cells (PBMCs), can serve as potential non-invasive biomarkers for NSCLC diagnosis, patient stratification, therapy monitoring, and prognosis. Methods: RNA was isolated from PBMCs of 136 NSCLC patients and 64 healthy donors. RT–qPCR quantified miR expression in PBMCs after predefined QC filtering: miR-155-3p (NSCLC n = 63; controls n = 28), miR-3196 (NSCLC n = 55; controls n = 28), and miR-155-5p (NSCLC n = 23; controls n = 12). Diagnostic performance was assessed using receiver operating characteristic (ROC) analyses, reporting area under the curve (AUC), and threshold-dependent sensitivity/specificity. Survival was analyzed with Kaplan–Meier/Cox methods. Associations with clinicopathological variables (stage, metastasis, smoking, EGFR, and KRAS status), treatment response (chemotherapy, immunotherapy, TKIs), and survival outcomes were examined. Results: miR-155-3p was upregulated in NSCLC, whereas miR-3196 was downregulated relative to controls; AUCs were 0.881 and 0.784, respectively. At high-sensitivity operating points, specificity was lower (≈29–30%), consistent with PBMC miRs reflecting both immune activation and tumor burden. In adenocarcinoma, miR-155-3p was associated with advanced stage, metastatic disease and smoking history. miR-3196 aligned with features of metastatic progression. During systemic therapy (chemotherapy, immunotherapy, TKIs), circulating levels of both miRs tended to normalize. Notably, normalization of miR-155-3p levels was associated with improved overall survival, supporting its prognostic value and utility for treatment monitoring. Conclusions: Circulating miR-155-3p and miR-3196 in PBMCs are promising screening/monitoring non-invasive candidates rather than stand-alone NSCLC diagnostics at current thresholds. Combining these miRs with additional biomarkers and/or clinical covariates and tuning decision thresholds may enhance specificity for diagnostic use. While preliminary, these findings warrant validation in large, prospective studies with standardized protocols to enable clinical implementation.