ARCHIVES

Financial markets are increasingly driven by a combination of quantitative trends and qualitative market sentiment originating from news, macroeconomic events, and social media discussions. Traditional financial forecasting systems that rely solely on historical time-series information often fail to capture sudden volatility caused by real-world events. This research presents AutoTraderX, a hybrid cloud-native financial intelligence framework integrating Agentic AI workflows, Retrieval-Augmented Generation (RAG), and Fine-Tuned Large Language Models for real-time explainable trading support. The proposed framework combines real-time market data ingestion, vector database retrieval, and contextual reasoning using a Quantized Llama-3 Large Language Model fine-tuned using QLoRA techniques. The system generates explainable BUY/SELL/HOLD recommendations while maintaining low inference latency and high semantic precision. Experimental evaluation on high-cap equities and cryptocurrencies demonstrates that AutoTraderX achieves significantly improved directional accuracy and risk-adjusted returns when compared with traditional forecasting baselines such as ARIMA, LSTM, and sentiment-only architectures. The framework also ensures scalability and fault tolerance using Kubernetes-based microservice deployment on AWS EKS infrastructure.