Jianan Tang
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Research Subjects
Jianan Tang has a total of 6 co-authors. Here are the top co-authors by number of shared publications:
- Qiudan Wang
Cisco Systems Inc., San Jose, CA, USA
2 shared publications Shubha Pant
Cisco Systems Inc., San Jose, CA, USA
2 shared publications- Mahesh Viswanathan
Cisco Systems, Inc., San Jose, CA, USA
1 shared publication Nihar Dandekar
Cisco Systems Inc., San Jose, CA, USA
1 shared publication- Shaja Arul Selvamani
Cisco Systems Inc., San Jose, CA, USA
1 shared publication - Zahra Aminiranjbar
Cisco Systems, Inc., San Jose, CA, USA
1 shared publication
Latest Publications
- Open Access
- 26 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 154142-154152
Smarter Root Cause Analysis: Enhancing BARO With Outlier Filtering and Ranking Refinement
Root Cause Analysis (RCA) is essential for diagnosing failures in modern software systems. BARO, which integrates Multivariate Bayesian Online Change Point Detection with a Robust Scorer, has demonstrated strong performance on RCAEval, a leading RCA benchmark for microservices. This method follows a two-stage process: it first detects failure onset and then ranks potential root causes by analyzing pre-failure and post-failure metric distributions. However, its sensitivity to extreme metric spikes can lead to unstable rankings and misdiagnoses. To address this, we introduce two key enhancements: (1) outlier filtering mechanisms to remove transient anomalies and (2) robust aggregation-based scoring to reduce the influence of fleeting spikes while emphasizing sustained deviations. Empirical evaluations on RCAEval show the enhanced BARO raises Avg@5 accuracy from 0.77 to 0.92 while adding only 0.21 s of extra processing time per failure case.
- 26 August 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 154142-154152
Open AccessSmarter Root Cause Analysis: Enhancing BARO With Outlier Filtering and Ranking Refinement
Root Cause Analysis (RCA) is essential for diagnosing failures in modern software systems. BARO, which integrates Multivariate Bayesian Online Change Point Detection with a Robust Scorer, has demonstrated strong performance on RCAEval, a leading RCA benchmark for microservices. This method follows a two-stage process: it first detects failure onset and then ranks potential root causes by analyzing pre-failure and post-failure metric distributions. However, its sensitivity to extreme metric spikes can lead to unstable rankings and misdiagnoses. To address this, we introduce two key enhancements: (1) outlier filtering mechanisms to remove transient anomalies and (2) robust aggregation-based scoring to reduce the influence of fleeting spikes while emphasizing sustained deviations. Empirical evaluations on RCAEval show the enhanced BARO raises Avg@5 accuracy from 0.77 to 0.92 while adding only 0.21 s of extra processing time per failure case.
- Open Access
- 11 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 138795-138812
DAWN: Designing Distributed Agents in a Worldwide Network
The rapid evolution of Large Language Models (LLMs) has transformed them from basic conversational tools into sophisticated entities capable of complex reasoning and decision-making. These advancements have led to the development of specialized LLM-based agents designed for diverse tasks such as coding and web browsing. As these agents become more capable, the need for a robust framework that facilitates global agentic communication and collaboration for building sophisticated software solutions has become increasingly important. Distributed Agents in a Worldwide Network (DAWN) addresses this need by providing an architectural framework that allows globally distributed agents of any provenance to be registered, discovered, and organized for building AI-based applications and solutions. In DAWN, a Principal Agent Service composes and oversees the execution of agentic applications. It delegates tasks to one or more Gateway Agent Services that provide for the discovery, registration, and connection of the most suitable agents to fit each application’s need. DAWN offers three operational modes: No-LLM mode for deterministic and classical software development, Copilot for decision-making augmented using AI, and LLM Agent for autonomous operations. Last but not the least, DAWN ensures the safety and security of agent collaborations globally through a dedicated safety, security, and compliance layer, protecting the network against attackers and adhering to stringent security and compliance standards. These features make DAWN a robust framework for designing, developing and deploying agent-based applications across business and consumer applications.
- 11 July 2025
- Published by Institute of Electrical and Electronics Engineers (IEEE) in IEEE Access
- Vol. 13, 138795-138812
Open AccessDAWN: Designing Distributed Agents in a Worldwide Network
The rapid evolution of Large Language Models (LLMs) has transformed them from basic conversational tools into sophisticated entities capable of complex reasoning and decision-making. These advancements have led to the development of specialized LLM-based agents designed for diverse tasks such as coding and web browsing. As these agents become more capable, the need for a robust framework that facilitates global agentic communication and collaboration for building sophisticated software solutions has become increasingly important. Distributed Agents in a Worldwide Network (DAWN) addresses this need by providing an architectural framework that allows globally distributed agents of any provenance to be registered, discovered, and organized for building AI-based applications and solutions. In DAWN, a Principal Agent Service composes and oversees the execution of agentic applications. It delegates tasks to one or more Gateway Agent Services that provide for the discovery, registration, and connection of the most suitable agents to fit each application’s need. DAWN offers three operational modes: No-LLM mode for deterministic and classical software development, Copilot for decision-making augmented using AI, and LLM Agent for autonomous operations. Last but not the least, DAWN ensures the safety and security of agent collaborations globally through a dedicated safety, security, and compliance layer, protecting the network against attackers and adhering to stringent security and compliance standards. These features make DAWN a robust framework for designing, developing and deploying agent-based applications across business and consumer applications.