Autonomous AI Trading in a Simulated Stock Market
Faculty Mentor
Tony Tian
Presentation Type
Poster
Start Date
4-14-2026 2:00 PM
End Date
4-14-2026 4:00 PM
Location
PUB NCR
Primary Discipline of Presentation
Computer Science
Abstract
This project presents a full-stack simulated stock market featuring 15,000 concurrent algorithmic trading bots and an autonomous AI trading agent. The backend is a multithreaded Java application hosted on AWS EC2 serving REST APIs, simulating real-time order book mechanics across 2,000 stock listings. A Next.js frontend deployed on Vercel provides a live dashboard for monitoring market activity. The core research component is an LLM-powered agent using Google Gemini's function calling API, which autonomously scans markets, analyzes company financials, and executes trades without human intervention. Key findings include the challenges of concurrent order matching in adversarial environments, architectural trade-offs in split cloud deployments, and the effectiveness of LLM tool use for autonomous multi-step decision-making in competitive real-time systems.
Recommended Citation
Layden, Carson and Helfenstein, Warren, "Autonomous AI Trading in a Simulated Stock Market" (2026). 2026 Symposium. 28.
https://dc.ewu.edu/srcw_2026/ps_2026/p3_2026/28
Creative Commons License
This work is licensed under a Creative Commons Attribution-NonCommercial-No Derivative Works 4.0 International License.
Autonomous AI Trading in a Simulated Stock Market
PUB NCR
This project presents a full-stack simulated stock market featuring 15,000 concurrent algorithmic trading bots and an autonomous AI trading agent. The backend is a multithreaded Java application hosted on AWS EC2 serving REST APIs, simulating real-time order book mechanics across 2,000 stock listings. A Next.js frontend deployed on Vercel provides a live dashboard for monitoring market activity. The core research component is an LLM-powered agent using Google Gemini's function calling API, which autonomously scans markets, analyzes company financials, and executes trades without human intervention. Key findings include the challenges of concurrent order matching in adversarial environments, architectural trade-offs in split cloud deployments, and the effectiveness of LLM tool use for autonomous multi-step decision-making in competitive real-time systems.
