I’ve built a high-performance, open-source C++ simulation engine designed to ingest streams of tokens (e.g., from large language models) and convert them into live trade signals on sub-10 microsecond latency.

The engine performs semantic mapping of tokens like “crash,” “bullish,” or “panic” into weighted market bias and volatility signals — which then dynamically adjust trading strategy parameters at fractional time scales.

It supports lock-free concurrency, zero-copy streaming, configurable sensitivities, and detailed logging for research and experimentation.

I made it as a pet project exploring the fusion of NLP and quantitative trading at low latency.

Would love to get feedback from the HN community, especially those with expertise in quant finance, low-latency systems, or AI-driven market strategies.

Here’s the repo and a sample token stream for quick testing.