Milad Sabouri
AI/ML Researcher & Engineer
Building Personalized & Intelligent Decision Systems
About#
I am an Applied Scientist and a Ph.D. candidate in Computer Science at DePaul University.
My current research in intelligent systems is grounded in my extensive industry background as a software engineer, which informs my approach to building robust, scalable, and maintainable models.
My current research and development centers on architecting LLM-powered AI agents for dynamic and explainable user modeling within recommender systems. Leveraging my foundation in reinforcement learning and deep learning, I build these systems to enhance personalization while ensuring a high degree of model transparency.
Applied Research & Engineering#
My work involves the research and development of intelligent systems designed for adaptivity, transparency, and efficacy in real-world environments. My approach is centered on three core areas of research and application:
Personalization & Recommender Systems#
My research in this area focuses on developing recommendation models with a more robust understanding of user preferences. I construct hybrid architectures that integrate both relational and contextual data to effectively model complex, temporal user behavior. This research also addresses the challenge of satisfying competing objectives within a recommendation platform, for which I develop multi-objective optimization algorithms to generate more balanced and effective outcomes.
Intelligent Decision & Optimization Systems#
My work in this domain involves engineering autonomous systems that learn and adapt in dynamic environments. Using reinforcement learning, I develop algorithms designed to optimize for long-term objectives by making sequential decisions based on user interaction data. This methodology is applicable to problems such as optimizing personalized incentive strategies and managing dynamic user engagement.
LLM-Powered AI Agents & Explainable AI#
My current research investigates the application of LLM-powered AI agents for user modeling. I am developing methods where these agents are utilized to construct sophisticated and dynamic representations of user behavior, with the goal of improving personalization. A core objective of this research is to advance Explainable AI (XAI). By designing agent prompts that elicit transparent, human-readable rationales for model outputs, I work towards systems that are both effective and inherently auditable, a central theme in my published research.
Publications#
Using LLMs to Capture Users’ Temporal Context for Recommendation 19th ACM Conference on Recommender Systems (RecSys ’25), Context-Aware Recommender Systems (CARS ’25), 2025.
Paper · Code
Towards Explainable Temporal User Profiling with LLMs. 33rd ACM Conference on User Modeling, Adaptation and Personalization (UMAP ‘25), Explainable User Models and Personalized Systems (ExUM ‘25), 2025.
Paper
Beyond Static Calibration: The Impact of User Preference Dynamics on Calibrated Recommendation. 32nd ACM Conference on User Modeling, Adaptation and Personalization (UMAP ‘24)., 2024.
Paper
Personalized Educational Learning with Multi-Stakeholder Optimizations. Fairness in User Modeling, Adaptation and Personalization (FairUMAP 2019), 2019.
Paper
I am always open to discussing new research, collaborations, or opportunities. Please feel free to get in touch.
Find me here:
