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 vLLM and 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.

View Architecture & Code on GitHub