AI Agent System Architect & Trainer — Self-Employed
Text
Prompt + Response Writing (SFT)Trained and optimized multiple AI agents using structured prompts and workflow orchestration for real-time A-share market monitoring and automated report generation. Implemented multi-layer system prompts to support complex multi-turn reasoning and accurate integration of multiple financial data sources. Evaluated and improved output quality with scoring, anomaly detection, and multi-source cross-validation to maintain low error rates and high task accuracy. • Data pipelines included data collection, cleaning, cross-validation across Sina, Tencent, and Citi Securities, analysis, and HTML/Markdown report generation • Labeled/structured training signals centered on finance domain tasks such as sector rotation prediction and event impact duration modeling • Built historical similarity retrieval workflows using 22+ indexed events for agent reference and response grounding • Applied >95% accuracy targets through iterative QA loops and cross-source validation