Conventional interview processes continue to exhibit persistent limitations, including subjective evaluations, susceptibility to bias, and limited scalability when assessing large candidate pools. These issues adversely affect both the fairness of candidate assessment and the operational efficiency of recruitment workflows. This paper introduces TalentSync, an AI-driven multi-agent interview platform designed to enable structured, adaptive, and scalable interview assessments. The system combines large language models with multimodal behavioral analysis to emulate realistic interview scenarios, interpret verbal and non-verbal cues, and generate adaptive feedback aligned with candidate performance. TalentSync employs a coordinated architecture comprising a Role Agent, Environment Agent, and Memory Agent, which operate in parallel to manage question generation, contextual reasoning, response evaluation, and feedback synthesis with minimal human intervention. In contrast to conventional automated interview systems that rely on sequential processing or single-model evaluation, the proposed platform adopts a parallel, pipelined execution strategy that enhances real-time adaptability while reducing system latency. Additionally, TalentSync supports dual-perspective evaluation by delivering actionable insights to candidates and comparative analytics to recruiters. Experimental evaluation and qualitative analysis demonstrate improvements in evaluation consistency, contextual relevance, and assessment transparency. Overall, the proposed system illustrates how structured multi-agent AI architectures can improve fairness, scalability, and decision-making in modern recruitment environments while complementing human judgment rather than replacing it.