PhD Research Interest

As a Data Scientist at HDR, a global professional services firm specializing in engineering and architecture, I have gained extensive experience implementing advanced Large Language Model (LLM) systems within organizational processes. A major challenge currently facing organizations is the “Black Box” nature of these systems. While most executive teams are enthusiastic about deployment, there is a lack of a robust methodology for detecting when agentic systems are underperforming or producing unreliable outputs.

For doctoral research, I propose developing an applied framework to quantify reliability and uncertainty within agentic LLM architectures. Unlike most existing research focused on single-shot LLM prompting and evaluation, I intend to investigate how uncertainty propagates through multi-step LLM frameworks such as LangChain and LangGraph. Using probabilistic modeling, calculus-based optimization and reinforcement learning, I strive to identify inflection points at which human review becomes necessary.

This applied research centers on examining how established observability metrics including output variability, self-consistency, and prompt sensitivity along with uncertainty optimization and reinforcement learning can support trustworthy agentic LLM deployments. My research objective is to develop a diagnostic layer that enables organizations to move beyond generic oversight policies toward a more data-driven “human-in-the-loop” strategy. My academic experience in Multivariate Statistics and Calculus and professional experience with LLM systems provides the skillset needed to conduct novel agentic uncertainty research.