Address:
05-129, Building 801, 1101 Kitchawan Rd
Yorktown Heights, NY 10598
About Me
Saurabh Jha is a senior research scientist at IBM Research. He works at the intersection of AI and systems, focusing on building resilient systems for AI agents and developing AI agents for IT operations. His research has been integrated into multiple IBM products and contributed to positioning Instana as a leader in 2025 Gartner’s Magic Quadrant for Observability Platforms.
Saurabh Jha is a senior research scientist at IBM Research in Yorktown Heights, NY. He earned his Ph.D. in 2021 and his M.S. in Computer Science in 2016 from the University of Illinois at Urbana-Champaign (UIUC), following his B.Tech from VIT University. Before joining UIUC, he was a visiting scholar at NTU Singapore and he has previously worked with IBM, NVIDIA Research, and Microsoft Research as an intern or consultant.
Saurabh works at the intersection of AI and systems, spanning the vertical stack from AI agents for IT management tasks to optimizing the system stack for efficient execution of AI agents. His research has contributed to the integration of several innovative features into products deployed at scale. Most notably, his work has significantly enhanced the analytics capabilities of IBM Instana and Turbonomic, improving the resilience and performance of IT applications. His contributions helped position Instana as a leader in 2025 Gartner's Magic Quadrant for Observability Platforms. Additionally, his work contributed to the improved resilience of NVIDIA's DriveAGX platform for self-driving cars.
Saurabh has received numerous accolades, including the IBM Tech Award in 2024, the IBM Ph.D. Fellowship (2020-2021), the SAP Industry Ph.D. Scholarship (2014), and the VIT University Best Outgoing Student Award (2014). His work has earned best paper awards and honorable mentions at prestigious conferences such as ACM/IEEE SC 2020, IEEE ISSRE 2020, and ACM HPDC 2013.
Research Focus Areas
Resilient and Performant AI Systems: Building robust and efficient systems for AI agents orchestration.
AI for Systems: Building AI agents for IT operations tasks.
Recent News
Released the Artificial Analysis Cyber Index in partnership with Artificial Analysis to evaluate AI models on enterprise cyber defense tasks. Cyber Index
(2026/08/25) ServiceNow uses ITBench in StarHarness to evaluate agent harnesses for enterprise IT operations. StarHarness Paper
Four new papers: "Explanations over Graphs: An Agent Architecture for IT Enterprise Diagnostic" (NeurIPS 2026); "Towards a Belief-Based World Model for LLM Agents" (NeurIPS 2026 Continual World Models Workshop); "The Sorry State of Stateful Scaling" and "Load Fiction: How Your Load Generator is Misleading You" (SoCC 2026). Publications
(2026/07/14) IBM announced the Proof of Exploitability Agent in Concert Protect, now available in Tech Preview. I served as chief architect of the feature. Announcement
(2026/06/24) Two papers accepted at DSN 2026! "Reliable Orchestration of LLM Programs via Deterministic–Generative Separation" (Disrupt Track) and "PRAXIS: Integrating Program Analysis with Observability for Root-Cause Analysis" (Best Paper Award).
Conducted fault injection on NVIDIA AV software. Results were used by architects to deliver the stack to high-visibility customers. DSN [2020]DSN [2018]
Integration of Store-pings in NCSA's Blue Waters SC [2020]
Awards
ITBench at ICML 2025 selected as a spotlight paper
