A Scalable AI-Driven Natural Language Interface for Algorithmic Trading with Strategy Validation and Asynchronous Execution

This paper presents an advanced AI-assisted trading platform designed to enable users to conceptualize, test, and activate cryptocurrency trading strategies using plain-language descriptions, without requiring programming knowledge. Traditional algorithmic trading solutions impose significant technical barriers—demanding proficiency in coding and quantitative finance—that systematically exclude a broad segment of individual investors. The proposed system addresses this challenge by incorporating a Large Language Model (LLM) capable of interpreting free-form user descriptions and converting them into structured, machine-readable trading instructions. The platform is built on a distributed, service-oriented architecture comprising a Flutter-based mobile frontend, a Django REST API for core application logic, and a FastAPI microservice dedicated to LLM inference. Computationally intensive workflows—including strategy generation, simulation over historical data, and trade order dispatch—are handled through a non-blocking task pipeline orchestrated by Redis and Celery. Prior to execution, each strategy is passed through a rule-based validation module that verifies logical integrity and compliance with defined risk parameters. A built-in backtesting engine enables retrospective performance analysis, and live trading is supported through persistent connections to cryptocurrency exchange APIs. Experimental results demonstrate that the platform substantially reduces the skill threshold required for strategy development, expands market participation for non-technical users, and maintains reliable throughput through distributed task handling. The modular, loosely coupled design supports horizontal scaling and straightforward integration with production financial systems