The growing number of fine-granular data streams opens up new opportunities for improved risk analysis, situation and evolution monitoring as well as event detection. However, there are still some major roadblocks for leveraging the full potential of data stream processing, as it would, for example, be needed the highly relevant systemic risk analysis in the financial domain.The QualiMaster project will address those road blocks by developing novel approaches for autonomously dealing with load and need changes in large-scale data stream processing, while opportunistically exploiting the available resources for increasing analysis depth whenever possible. For this purpose, the QualiMaster infrastructure will enable autonomous proactive, reflective and cross-pipeline adaptation, in addition to the more traditional reactive adaptation. Starting from configurable stream processing pipelines, adaptation will be based on quality-aware component description, pipelines optimization and the systematic exploitation of families of approximate algorithms with different quality/performance tradeoffs. However, adaptation will not be restricted to the software level alone: We will go a level further by investigating the systematic translation of stream processing algorithms into code for reconfigurable hardware and the synergistic exploitation of such hardware-based processing in adaptive high performance large-scale data processing. The project focuses on financial analysis based on combining financial data streams and social web data, especially for systemic risk analysis. Our user-driven approach involves two SMEs from the financial sector. Rigorous evaluation with real world data loads from the financial domain enriched with relevant social Web content will further stress the applicability of QualiMaster results.