This project addresses the limitation of Large Language Models (LLMs) in performing autonomous, cross-application tasks. We developed an Autonomous AI Agent within the AppWorld environment capable of Long-horizon planning. Unlike standard chatbots, this agent translates natural language instructions into executable API calls across multiple simulated interfaces (e.g., Email, Banking, Calendar). The system employs a dynamic reasoning engine and a feedback-driven control loop to execute, verify, and adapt actions, ensuring the reliable completion of complex, multi-step tasks.

LINKS:

INITIAL IDEA DOCX:
https://docs.google.com/document/d/1wEN4Vp_tdEE70-Z_fxhgjp2CfqkeBaQGvedPbANsu0Q/edit?tab=t.0

HIGH-LEVEN DESIGN:
https://docs.google.com/document/d/1ewSbDkUHpcBb4wjJsOqbFOs-XZk2XqmJkA3mLapHRg4/edit?usp=sharing