Advances in large language models (LLMs) have unlocked new possibilities in financial sentiment analysis and market movement prediction. We present a fine-tuned Falcon-1B model trained on the Financial Phrasebank dataset and further enhanced with In-Context Backtranslation (IBT), a method that improves generalization by rephrasing text through translation. Fine-tuning is done using Quantized Low-Rank Adaptation (QLoRA), a memory- and compute-efficient method that updates only a small set of parameters. Our model achieves 91% accuracy on a 200-sample test set and performs well at scale, outperforming existing instruction-tuned and domain-specific models. To demonstrate real-world value, we apply the model to analyze sentiment-driven industry movements. This pipeline offers an efficient, scalable way to adapt general-purpose LLMs for financial applications.