Named Entity Recognition represents a crucial component in natural language processing (NLP), involving the identification and categorization of named entities, such as persons, organizations, and locations, within textual data. Despite significant advancements in NER for prominent languages like English and Arabic, recognizing named entities within specific dialects presents distinctive challenges. Algerian Arabic, a widely spoken variant with unique linguistic characteristics, has garnered limited attention in NER research compared to standardized Arabic varieties. This study introduces a novel architecture that addresses the NER challenges specific to Algerian Arabic. Our proposed framework integrates the Multi-dialect BERT model, Bidirectional Gated Recurrent Units (Bi-GRU), and Conditional Random Field (CRF) layer. Our experimental findings showcase promising results across all architecture variants. Particularly, the Multi-dialect-Arabic-BERT-Bi-GRU-CRF model emerges as the most effective, achieving a remarkable 93.04% F1-score. This success underscores its proficiency in capturing sequential dependencies and decoding complex linguistic structures inherent in Algerian Arabic text.