Generative AI Powers Autonomous Networks: A PhD Research Deep Dive
Shaghayegh (Shirley) Shajarian, a third-year PhD student in Computer Science at North Carolina A&T State University, is conducting pioneering research on the application of generative AI to computer networks. Working under the guidance of Dr. Sajad Khorsandroo and Dr. Mahmoud Abdelsalam within the Autonomous Cybersecurity and Resilience Lab, her work aims to significantly reduce manual labor in network management and pave the way for fully autonomous, self-running networks.
Shajarian's research primarily focuses on developing AI-driven agents designed to automate critical network operations. These agents are being engineered to assist with tasks such as analyzing network logs, troubleshooting technical issues, and generating comprehensive documentation. The overarching vision is to enable networks to autonomously configure, optimize, heal, and protect themselves, thereby minimizing human intervention. Beyond this core area, Shajarian also has a keen interest in network cybersecurity, evidenced by her published studies on classifying malicious domains using transfer learning and a survey of explainable AI (XAI) techniques for malware analysis.
A particularly compelling aspect of her work involves leveraging Large Language Models (LLMs) as intelligent agents within network systems. As modern computer networks grow increasingly complex, manual management is becoming both costly and unsustainable. Shajarian finds the potential of LLMs fascinating because of their ability to interpret vast amounts of log data, accurately identify issues, and communicate their findings in a human-like conversational manner. This capability is crucial for supporting network operators and facilitating the transition to semi-autonomous networks, where LLM-driven agents handle routine tasks while humans maintain oversight for verification and critical decision-making. Ultimately, this research brings closer the prospect of fully autonomous networks that can operate, adapt, and respond without constant human input.
Looking ahead, Shajarian plans to expand her study by integrating real-world telemetry and network logs. This will enhance situational awareness and support more effective decision-making in dynamic network environments. Her future work will also explore the practical deployment of LLM-based agents in real-world settings, focusing on their reliability and adaptability to changing network conditions. A key objective is to evaluate how these systems can autonomously identify, diagnose, and document network issues, while always ensuring that human oversight remains in place for critical decision points.
Shajarian's journey into AI, particularly LLMs, began during her undergraduate and master’s studies in computer software engineering, where she became intrigued by how machine learning models could mimic human reasoning. The rapid advancements in LLMs further fueled her curiosity, especially regarding their application in complex domains like computer networks. She believes their capacity to foster system autonomy is vital for alleviating the burden on operators and reducing the need for manual intervention.
For individuals considering a PhD in this interdisciplinary field, Shajarian emphasizes the importance of flexibility, curiosity, and staying abreast of the latest developments. A PhD combining AI and computer networks demands depth in both areas, requiring discipline to balance diverse technical foundations. She advises aspiring researchers to choose an advisor and a research topic that genuinely aligns with their interests, as this alignment provides essential motivation during challenging phases. Ultimately, she encourages pursuing a PhD not merely for the degree, but for a deep-seated desire to ask questions, find answers, and contribute meaningfully to the field.
Outside of her groundbreaking research, Shajarian is a passionate cook, aspiring to earn a Michelin star. She views cooking as her creative outlet and a favorite way to connect with others.