Yahya Al-Eryani

Conducting research on privacy risks in foundation model inference focusing on identifying and mitigating inference-time side-channel vulnerabilities. Designing a security analysis framework for evaluating prompt leakage through timing and cache-based attack vectors in LLM serving systems. Proposing and assessing practical defense mechanisms to ensure secure and privacy-preserving inference without compromising performance. Contributing to the advancement of secure AI serving strategies within multi-tenant and production-scale environments.

Developed large language models and generative AI solutions aimed at streamlining processes in diverse business domains and applying these models to real-world applications, enhancing AI model accuracy and user interface. Administered Azure cloud environments optimizing for software application protocols, scalability, and robust model governance. Managed entire model development lifecycle from data analysis and rule-based prototyping to mature machine learning product development. Documented models and supported model governance ensuring validation and explainability of AI models.

Developed advanced machine learning models focusing on generative AI for on-premises and private cloud environments. Worked closely with infrastructure and application teams to define requirements and deliver AI solutions. Engineered microservices to support AI functionalities ensuring robust and scalable performance. Implemented LLMs, vector embeddings, and knowledge graphs while leading testing and validation of models using CI/CD practices.