I work on LLM evaluation, agentic systems and AI SRE: making AI systems, and the infrastructure under them, measurable, debuggable and trustworthy in production.
Previously Staff Research Scientist at IBM Research. Ph.D. in Computer Science, NC State University, advised by Xiaohui (Helen) Gu.
InsightFinder's predictive reliability agent. It predicts failures hours before impact, identifies root causes and resolves issues across the stack, from AI agents and applications to IT infrastructure, with engineer approval.
Evaluates LLM and agent responses for hallucination, relevance, safety and bias, guards against jailbreaks, prompt injection and PII leakage, and compares models and prompt versions side by side before they ship.
Benchmarking agents on real-world SRE, compliance and FinOps tasks (ITBench); generating, assessing and refining operations scripts (ScriptSmith); validating generated code without references (CodeSift).
Distributed concurrency bugs through RPC-driven, race-directed test generation and fuzzing (Themis); server-side request forgeries through LLM-assisted, path-sensitive taint analysis (Artemis).
Detecting risky scripts embedded in Ansible infrastructure code (SecureCode); how well static and dynamic schemes catch container exploits; privacy-preserving, auditable smart contracts on Hyperledger Fabric (FabZK).
Hybrid static and runtime diagnosis, data-corruption hang detection, automatic hang and timeout fixing. The subject of my Ph.D. dissertation.