Date of Award
10-2025
Culminating Project Type
Starred Paper
Styleguide
ieee
Degree Name
Computer Science: M.S.
Department
Computer Science and Information Technology
College
School of Science and Engineering
First Advisor
Mark Petzold
Second Advisor
Bilal Al-Ahmed
Third Advisor
Akalanka Mailewa
Creative Commons License
This work is licensed under a Creative Commons Attribution-Noncommercial-No Derivative Works 4.0 License.
Keywords and Subject Headings
Robotic Process Automation, XAML Code Generation, Reinforcement Learning, UiPath, Intelligent Automation, PEFT, multimodal model, Visual Language Pre-training Model (VLP)
Abstract
Robotic Process Automation (RPA) workflows generated to automate repetitive tasks are vulnerable to failure and often incur high maintenance costs due to the ever-evolving nature of client applications and graphical user interfaces (GUIs). To address this, we propose embedding large language models (LLMs), fine-tuned using Parameter-Efficient Fine-Tuning (PEFT), into the workflow generation and validation pipeline to generate and update XAML snippets based on user-defined requirements. Together with a memoryless and stateless sandbox testing environment, workflows can be autonomously and securely validated, leveraging Reinforcement Learning (RL) for iterative improvement. This paper presents a conceptual design of such a system and explores its potential, as well as its inherent limitations.
Recommended Citation
Waqar, Maliha, "Self-Correcting Automated Workflows with XAML Code Generation" (2025). Culminating Projects in Computer Science and Information Technology. 61.
https://repository.stcloudstate.edu/csit_etds/61
