Tanzim Mahfuz

I am a Ph.D. candidate in Electrical and Computer Engineering at the University of Maine, where my research focuses on building explainable, trustworthy, and resilient AI systems for complex real-world problems. During my Ph.D., I have worked across hardware security, AI security, digital design automation, computer vision, and agentic AI, developing complete research pipelines from problem formulation and data generation to model design, implementation, evaluation, and publication. I led X-DFS, an explainable AI framework for security-aware design-space exploration published in IEEE TIFS, and POLARIS, a machine learning framework for power side-channel mitigation published at DAC. I also co-developed DASH, a differentiable adversarial machine learning framework accepted at CVPR 2026, and have contributed to multi-agent AI research, including systems for tool use, planning, verification, and the evaluation of harmful behavior across interacting agents.

Alongside my publications, I developed and provisionally filed two patents involving generative and agentic AI for semiconductor design. One focuses on generative AI-based leakage-aware secure hardware transformations, while the other introduces an explainable multi-agent framework for evaluating cross-PDK migration difficulty across multiple technical domains. These projects gave me experience taking an early research idea through system design, implementation, evaluation, collaboration, and intellectual property development.

I am particularly interested in LLM architecture because I believe the architectural principles behind modern language models are reshaping how intelligent systems learn representations, reason across multiple steps, use tools, retain context, and interact with users. I want to understand these systems beyond surface-level application and contribute to the design of models and agentic systems that are reliable, adaptable, and useful in practice. Before beginning my Ph.D., I worked as a full-stack software engineer, which shaped my interest in building systems that are not only technically strong but also scalable, maintainable, and usable by others. The broader goal of my Ph.D. has been to become a researcher and engineer who can enter an unfamiliar domain, understand its core challenge, and independently develop an effective AI-based solution while communicating the work clearly to both technical and non-technical audiences.