Akila Balasubramanian

AI Agents That Investigate, Not Just Predict

Abstract:

Modern distributed systems generate high-cardinality telemetry across metrics, logs, and traces, resulting in a combinatorial search space for diagnosing failures. Root cause analysis (RCA) in such environments is inherently iterative, hypothesis-driven, and dependent on correlating heterogeneous signals under uncertainty. Existing approaches—including manual investigation, rule-based automation, and prompt-driven large language model (LLM) systems—fail to reliably support this process due to the absence of structured reasoning, persistent state, and deterministic execution. This paper presents an agentic systems framework for automated RCA, where troubleshooting is formulated as a closed-loop decision process over telemetry. A layered architecture is introduced, comprising a control layer for orchestration, a memory layer for token-aware context management, a tooling layer for deterministic interaction with observability systems, and a governance layer for enforcing correctness, security, and runtime observability. RCA is modeled as iterative hypothesis refinement, supported by algorithms for action selection, evidence aggregation, conflict resolution, and failure recovery. An empirical evaluation over 1,200 production-style troubleshooting tasks demonstrates that the proposed system improves task success rates from 61.8% to 86.7%, reduces user intervention by 3.5×, and decreases effective time-to-resolution by approximately 42%. Token consumption is reduced by up to 4.8× via adaptive memory strategies without loss of accuracy. Robustness experiments show nearly 2× improvement in failure recovery and significant gains in handling ambiguous inputs. These results indicate that agentic architectures transform observability from passive monitoring into active, automated reasoning systems, enabling scalable, reliable, and evidence-driven RCA.

Profile:

Akila Balasubramanian is a Software Engineering Technical Leader at Cisco, where she leads the design and development of AI-powered observability and intelligent troubleshooting capabilities for cloud-native platforms. Her work spans distributed systems, agentic AI, observability, and platform engineering, with a focus on building production-ready AI systems that help engineering teams detect, investigate, and resolve complex operational issues more effectively.

Over the years, Akila has led the development of technologies including AI-directed troubleshooting, automated root cause analysis, real user monitoring, session replay, telemetry intelligence, and intelligent investigation workflows. Her work emphasizes trustworthy AI, explainable reasoning, robust evaluation, and scalable platform architecture that can support enterprise-scale production environments.

Akila is passionate about bridging research and real-world engineering, translating advances in AI into reliable systems that solve meaningful customer problems. She frequently shares practical lessons on building agentic AI platforms, LLM evaluation, observability, and the engineering challenges of deploying intelligent systems at scale.

Download profile as PDF: Click Here