From Signals to Solutions: Grounded LLM Agents for Root Cause Analysis in Complex Systems

Speaker: Professor Hans-Arno Jacobsen
Electrical and Computer Engineering and Computer Science
University of Toronto
  and
Visiting Professor
Department of Computer Science and Engineering
HKUST

Title: From Signals to Solutions: Grounded LLM Agents for Root Cause Analysis in Complex Systems

Date: Monday, 2 November 2026

Time: 4:00pm - 5:00pm

Venue: Lecture Theater F (Leung Yat Sing Lecture Theater), near lift 25/26, HKUST

Abstract:

Modern cloud and microservice systems generate vast streams of heterogeneous telemetry—metrics, logs, and traces—yet turning this data into reliable diagnoses and actionable remediation remains a fundamental challenge. Existing approaches either operate on isolated signals, rely on unconstrained LLM reasoning prone to hallucination, or stop at fault ranking without closing the loop to resolution.

This talk presents a new paradigm for operational intelligence: graph-grounded, agentic AI for root cause analysis (RCA). We introduce a novel framework that combines causal inference over service dependency graphs with structured, graph-guided LLM reasoning, integrating multi-modal evidence—including a novel trace-based scoring mechanism—to produce accurate diagnoses, concise incident summaries, and actionable recommendations. Our results demonstrate substantial gains in diagnostic precision and actionability, pointing toward a new generation of trustworthy, graph-aware AI systems for operating complex distributed systems.

Biography:

Arno Jacobsen is a Professor of Electrical and Computer Engineering and Computer Science at the University of Toronto, where he holds the endowed Jeffrey Skoll Chair in Computer Networks and Innovation and is a Tier 1 Canada Research Chair in Data-Intensive Systems. He is an internationally renowned scholar in distributed systems and data management, particularly known for his groundbreaking contributions to distributed event-based and publish/subscribe systems, widely used in enterprise and Internet-scale settings. His research further spans AI systems infrastructure and AI-driven systems, addressing both the engineering of scalable systems for AI workloads and the integration of AI into large-scale system design (cf. msrg.org). He also conducts research on quantum computing, especially distributed quantum computing, and has led the Quantum Software Consortium of Canada (qscc.ca) as its Scientific Director. He has held leadership roles in major international conferences and is widely published in top IEEE/ACM venues. He has received Best Paper and Test-of-Time awards and is a Fellow of the IEEE and a Fellow of the Canadian Academy of Engineering.