Sarah.

Hi! I completed my Master's in Computer Science at Columbia University, where I was in the Software Systems track advised by Prof. Bjarne Stroustrup. I primarily worked with Prof. Gail Kaiser on hallucination in LLM-generated code and the alignment and reliability of deployed agents, and I was also part of the Complex Resilient Intelligent Systems Lab (CRIS) with Prof. Venkat Venkatasubramanian.

My research focuses on AI for code, hallucination, and alignment. I'm most interested in what happens when language models leave the benchmark and meet the real world: where they quietly make things up, how that failure shows up in generated code and in software engineering workflows, and what it takes to keep deployed agents behaving the way we intended. Recent work of mine studies how personality, base model, and guardrail specification shape the behavior of AI agents deployed in a live social network, and whether language models can reason about legal doctrine well enough to be trusted in enforcement settings.

Before Columbia, I completed my Bachelor's of Computer Science at the University of Waterloo, where I was lucky to work across three labs: CrySP (Cryptography, Security, and Privacy), the Data Systems Group, and the DIGR Lab (Diversity and Intergroup Relations). Along the way I interned twice at Google — first on Gemini Enterprise and Vertex AI Search, then in AI and Infrastructure — and at Microsoft on Azure Networking, with earlier internships in banking and cybersecurity.

I'm always happy to talk about research or collaborate — feel free to reach out!