MarketEPOLL is a fast-response trading intelligence workflow built for event-driven market behavior, especially around earnings and major headlines. The system ingests newly published financial news, processes the text with low-latency Llama via Groq LPUs, and outputs a sentiment confidence score intended to estimate directional market impact.

On the engineering side, the pipeline combines RSS and webhook ingestion for sub-second capture, structured analysis components for consistency, and vector-backed context retrieval for better signal quality. The end result is a practical framework for turning real-time narrative information into actionable trading context rather than raw headline noise.

Technologies: Groq API, Llama 70B, OpenAI API, LangChain, Snowflake, TypeScript