An enterprise-grade event-driven quantitative trading engine powered by domain-adapted Large Language Models (FinLLM) for institutional sentiment arbitrage.
Core Architecture:
- Sub-Second NLP Parsing: High-throughput streaming inference using
vLLMand TensorRT-LLM to decode central bank announcements within milliseconds. - Orthogonal Alpha Generation: Converts unstructured semantic shifts into quantitative z-score signals orthogonal to traditional equity momentum factors.
- Dynamic Hedging: Automated portfolio rebalancing execution engine interfacing via FIX API.