Building an Autonomous Alpha Engine with AI

The article describes the development of Nous Ergon: Alpha Engine, a fully autonomous trading system that combines AI-driven research, quantitative prediction, and rule-based execution to generate sustained market alpha.

Details

The article explores the thesis behind Nous Ergon: Alpha Engine, a fully autonomous trading system that combines AI-driven research, quantitative prediction, and rule-based execution to generate sustained market alpha. The system is designed to overcome the limitations of using a single large language model (LLM) for trading by splitting the problem into three specialized layers: research, prediction, and execution. The research module uses LLM agents to assess the fundamental attractiveness of stocks based on news sentiment and analyst research, while the prediction module employs machine learning models to evaluate short-term technical timing. The execution module then applies deterministic risk rules to ensure the system's survival. The key design decision is to use LLM agents only in the research module, which allows the other components to run unlimited simulations and backtests without incurring the cost of LLM API calls. The article also discusses the concept of 'horizon separation', where the research module focuses on long-term fundamental attractiveness, while the prediction module handles short-term technical timing.