The stock market serves as a fundamental pillar of the global financial ecosystem, influencing macroeconomic stability and investment strategies. Among various financial indicators, the closing price is a critical metric, encapsulating aggregated market sentiment and informing risk assessment, portfolio optimization, and algorithmic trading. Despite its significance, precise forecasting of closing prices remains a formidable challenge due to the stochastic nature of financial markets, characterized by high volatility, non-stationarity, and susceptibility to exogenous economic shocks. We propose FinStock-Net, a multi-scale temporal fusion model that leverages Bidirectional Long Short-Term Memory (BiLSTM) networks for temporal analysis and a gated fusion mechanism to balance short-term (1-day), mid-term (7-day), and long-term (15-day) market dynamics, aligning with real-world decision-making. Our framework also integrates volatility-sensitive indicators, called VIX index, to enhance robustness against abrupt fluctuations. Unlike conventional approaches. FinStock-Net employs adaptive gating mechanisms to dynamically reweight temporal features, ensuring contextual alignment with real-world decision-making horizons. We benchmark FinStock-Net on publicly available financial datasets like Nifty50, Sensex, and S& P500, demonstrating its superior predictive accuracy over some existing models and establishing it as a robust framework for stock market forecasting. The code for our proposed model is available on Github.