Experience
Work, education, teaching, and service
Work
Current role
Senior Staff Data Scientist
Apr 2026 – now
- Earlier at InsightFinder AI
- Staff Data ScientistMar 2025 – Apr 2026
- Software Engineer InternJun 2016 – Aug 2016
- Lead the development of ARI, a multi-agent predictive AI SRE agent: a coordinator agent orchestrates specialist sub-agents for incident prediction, anomaly detection, root cause analysis, code analysis and patch fixing, Kubernetes deployment, Jira tickets and Slack/email notifications, with human-in-the-loop approval.
- Designed ARI's model-agnostic agent stack: LangGraph agents with tool use over MCP, per-agent LLM choice across OpenAI, Anthropic, Google and AWS Bedrock, live streaming, source citations and OpenTelemetry tracing.
- Built a RAG-grounded, claim-level LLM-as-a-judge (98% accuracy) for multi-agent workflows: each claim is verified against evidence from vector search, with deterministic number, date and rule checks first, on a self-hosted small language model without fine-tuning.
- Developed AI Evaluation & Guardrails: hallucination, relevance, safety and bias evaluation, jailbreak, prompt-injection and PII guardrails, RAG/GraphRAG answer evaluation, and statistically gated A/B releases of judge versions.
- Staff Research ScientistJul 2022 – Mar 2025
- Research Staff MemberSep 2019 – Jul 2022
- Graduate InternMay 2018 – Aug 2018
- Contributed to ITBench (ICML 2025 spotlight and oral), an open benchmark for AI agents on real-world SRE, compliance and FinOps tasks; implemented benchmark scenarios.
- Built LLM tools for SREs: ScriptSmith (AAAI'25), retrieval-augmented Bash script generation with self-refinement, and CodeSift (CLOUD'24), reference-free validation of LLM-generated code.
- Co-developed the Kestrel threat-hunting pipeline with STIX-shifter connectors (IBM TCAP Significant Contributor) and integrated Keylime remote attestation with IBM Cloud Satellite and Red Hat CoreOS.
- Developed SecureCode (SoCC'20), deployed in IBM Cloud's DevOps pipeline to catch risky scripts in infrastructure code, and co-developed multi-cloud compliance and vulnerability risk ranking integrated with IBM AIOps (3× IBM Research Accomplishment).
- Research AssistantMay 2015 – Jul 2019
- Built performance-bug detection and fixing systems for cloud servers, Hytrace (SoCC'17, TPDS'19), DScope (SoCC'18) and HangFix (SoCC'20), combining static code analysis with ML-based anomaly detection on system-call traces.
- Co-developed timeout-bug detection and fixing, TScope (ICAC'18) and TFix (ICDCS'19), and a study of real-world timeout problems (IC2E'18, best paper nominee).
Education
- North Carolina State University
- Ph.D. in Computer ScienceAdvisor: Xiaohui (Helen) Gu · Dissertation: A Hybrid Approach to Cloud System Performance Bug Detection, Diagnosis and Fix2014 – 2019
Nanjing University of
Posts and TelecommunicationsChina · 2007 – 2014
Posts and TelecommunicationsChina · 2007 – 2014
- Nanjing University of Posts and Telecommunications
- Master's in Computer Software and Theory2011 – 2014
- Bachelor's in Information Security2007 – 2011
Teaching
- Teaching Assistant · CSC 724 Advanced Distributed SystemsSpring 2019
- Teaching Assistant · CSC 226 Discrete MathematicsFall 2014, Spring 2015
Mentoring
- Seyyed Ali AyatiIBM-TAMU, NSF RINGs programAug 2023 – Oct 2023
- Bekir TurkkanIBM internshipMay 2021 – Jan 2022
- RunXiang (Sam) ChengIBM-UIUC programMay 2020 – Aug 2020
- Pavithra IyerDANCE labFall 2017
Academic service
Track chair
IEEE International Conference on Cloud Engineering (IC2E), Industry Track2024
Program committee
Editorial board
Journal of Systems Research (JSys)2021 – now
Journal reviewer
Panelist
Workshop on Hot Topics in Operating Systems (HotOS)2021