<p>This study presents FinDeep-SocNet, a novel hybrid deep learning framework designed to evaluate the socio-economic development of customers through public sector banking interactions. The framework integrates multiple AI modules, combining transactional behavior analysis with advanced predictive modeling of financial capacity to build a comprehensive socio-economic profile of each customer. The system incorporates three primary stages: (i) Transaction Classification using a fine-tuned FinBERT model to label banking transactions into expenditure/income categories, (ii) Cash Flow Prediction through a Bi-GRU (Bi-directional gated recurrent unit) based temporal module that learns from sequential financial activities, and (iii) Socio-Economic Profiling using a Transformer-based fusion module that consolidates demographic, transactional, and predicted financial features into a dense representation. These embeddings are then classified into development tiers (underdeveloped, developing, and developed) using a fully connected neural layer. The model is trained and evaluated on a synthesized dataset based on realistic public banking records and socio-financial labels. Performance is measured across multiple metrics including Accuracy, Precision, Recall, F1-Score, and AUC reported in the result section. Evaluation of experimental results showed overall accuracy of 93.8% and Precision of 92.1%. The framework helps identifying customer development levels, tailor financial services, and promote inclusive growth. This highlights the potential of AI in bridging the gap between banking infrastructure and socio-economic upliftment through data-driven decision-making.</p>