About

Ravishankar K. Iyer is the George and Ann Fisher Distinguished Professor of Engineering at the University of Illinois at Urbana-Champaign, with joint appointments in the Departments of Electrical and Computer Engineering and Computer Science. He holds positions in the Coordinated Science Laboratory, the National Center for Supercomputing Applications, the Carle Illinois College of Medicine, and the Carl R. Woese Institute for Genomic Biology. His research focuses on reliable and secure networks and systems, measurement and modeling, dependability and security validation, data analytics, computing for genomics research, machine learning, artificial intelligence, and the trustworthiness of AI systems and algorithms. He leads the DEPEND Group at CSL, which develops multidisciplinary systems and software combining deep measurement-driven analytics and machine learning, applied in domains such as trust, resilience, security of critical infrastructures, and health analytics for personalized medicine. Professor Iyer directs the Illinois/Mayo NSF Center for Computational Biotechnology and Genomic Medicine. He is a Fellow of AAAS, IEEE, and ACM, and has received numerous awards including the IEEE Emanuel R. Piore Award and the 2011 Outstanding Contributions award by ACM. His academic career includes leadership roles such as Director of the Coordinated Science Laboratory, interim Vice Chancellor for Research, and founder and chief scientist of Armored Computing, Inc. His extensive research contributions span dependable computing, security, systems, and AI, with a notable emphasis on measurement-driven analytics and applications in genomics and health.

Research topics

• Computer Science
• Artificial Intelligence
• Intensive care medicine
• Distributed computing
• Operating system
• Statistics
• Internal medicine
• Telecommunications
• Medicine

Selected publications

• Search and Retrieval in Dermatology Atlases of Histopathology Images for Risk Stratification of Cutaneous Squamous Cell Carcinoma

  medRxiv · 2026-01-06

  articleOpen access

  Abstract Cutaneous squamous cell carcinoma (cSCC) poses significant clinical challenges due to its rising incidence and potential for metastasis. Histopathologic risk stratification is further limited by substantial inter-observer variability. Unsupervised AI approaches based on content-based image retrieval offer scalable and interpretable decision support for diagnostic pathology. The objective of this study was to evaluate the use of image retrieval within histopathology atlases to stratify cSCC tumor differentiation from whole-slide images (WSIs), while comparing different patch selection and feature extraction strategies. This retrospective study included 552 archived WSIs comprising 385 well-differentiated, 102 moderately differentiated, and 66 poorly differentiated cases collected across Mayo Clinic sites in Arizona, Florida, and Minnesota. Image atlases were constructed using multiple patch aggregation strategies (Mosaic, Collage, and Montage) and deep learning encoders (KimiaNet, PathDino, and H-Optimus-0). A leave-one-WSI-out evaluation framework was used to assess differentiation classification performance using accuracy, specificity, sensitivity, and F1 score. Mosaic combined with KimiaNet achieved the highest Top-1 accuracy (74.9%) and specificity (92.6%), while Mosaic with H-Optimus-0 yielded the best Top-5 accuracy (79.0%) and macro-F1 score (62.6%). Collage combined with KimiaNet produced the highest Top-5 specificity (99.5%). The generalizability of the evaluated AI models varied across hospitals, reflecting differences in imaging protocols, staining practices, and patient populations. Overall, unsupervised image search and retrieval provides effective, annotation-free support for cSCC differentiation and has the potential to enhance dermatopathology workflows when appropriate combinations of patch selection and feature ex-traction methods are employed.

  PublisherOA PDFDOI

• Control Signals in Closed-Loop Spinal Cord Stimulation in Patients with Chronic Pain: A Scoping Review

  Neuromodulation Technology at the Neural Interface · 2026-01-01

  articleOpen access

  INTRODUCTION: Spinal cord stimulation (SCS) provides significant relief for patients with chronic pain; however, many approaches have limitations in programming complexity, personalization, and long-term efficacy. Traditional open-loop systems require manual programming and fail to adapt to patient-specific anatomical or physiological changes over time. In response, closed-loop SCS systems have emerged, offering real-time modulation based on biomarkers such as position and evoked compound action potentials (ECAPs). However, these systems still largely fail to integrate subjective aspects of pain alongside objective neural biomarkers. OBJECTIVES: The purpose of this scoping review is to evaluate the control signals and algorithms used by closed-loop SCS devices and identify directions for improving their efficacy. MATERIALS AND METHODS: Following Preferred Reporting Items for Systematic Reviews and Meta-Analyses for Scoping Reviews guidelines, the PubMed, SCOPUS, and Web of Science data bases were queried on December 14, 2024. Peer-reviewed studies written in English related to closed-loop SCS were included. The inclusion criteria were 1) SCS therapy for chronic pain, and 2) real-time modulation of stimulation parameters. The exclusion criteria included review studies, book chapters, conference proceedings, small animal studies, or works unrelated to chronic pain. Initially, 688 unique articles were identified. After screening by two independent reviewers, 28 articles met all the established criteria, encompassing 19 unique studies. RESULTS: Three studies investigated subjective states, such as rating of pain, mood, and paresthesias; seven used objective features, including position and movement, and nine studies incorporated ECAP characteristics as a control signal. To our knowledge, no existing model has fully integrated both subjective and biophysical markers to inform closed-loop stimulation parameters. CONCLUSIONS: A closed-loop SCS algorithm that incorporates subjective and objective features may hold potential to improve quality of life in patients with chronic pain. Combining these approaches in a temporally resolved manner, for example, integrating patient reports with continuous electrophysiologic information using a state space mathematical model, could allow more optimized and patient-specific SCS programming.

  PublisherOA PDFDOI

• 308 Personalized Modeling to Optimize Closed-Loop Spinal Cord Stimulation in Patients With Chronic Pain: A Scoping Review and Proposed Model

  Neurosurgery · 2026-03-26

  article
  PublisherDOI

• DSN 2026 Artifact Evaluation for Paper: PRAXIS: Integrating Program Analysis with Observability for Root-Cause Analysis

  Zenodo (CERN European Organization for Nuclear Research) · 2026-03-27

  otherOpen accessSenior author

  Artifact evaluation repository for paper: PRAXIS: Integrating Program Analysis with Observability for Root-Cause Analysis

  PublisherDOI

• Bacterial clusters are associated with the risk of severe disease progression in inflammatory bowel disease irrespective of conventional disease categories

  Microbiome Research Reports · 2026-03-18

  articleOpen access

  Background: Inflammatory bowel diseases (IBDs) are complex conditions marked by chronic inflammation in the gastrointestinal tract. Traditional classification separates IBD into Crohn’s disease and ulcerative colitis, but this division may not fully capture disease heterogeneity. Here, we examine whether microbiome-driven subtyping can describe novel clinical IBD phenotypes. To achieve this, we applied unsupervised clustering to fecal microbiota profiles from the population-based Inflammatory Bowel Disease in South-Eastern Norway III (IBSEN III) cohort. Methods: A Gaussian Mixture Model (GMM) was used to cluster participants with IBD based on microbiome composition and examine associations between clusters and clinical outcomes, including inflammatory markers and disease severity during the first year after inclusion. Results: Three microbiome-based clusters were identified: CLO (dominated by Clostridia UCG-014), ALF (Agathobacter, Lachnoclostridium, and Faecalibacterium), and RUM (Ruminococcus gnavus). Participants in the RUM cluster had a higher risk of future severe disease than those in the CLO cluster, even among participants with remission-to-mild disease at inclusion (21% vs. 6%, P < 0.00001). This association could not be explained by antibiotic use or baseline disease severity. Cluster membership alone performed comparably to fecal calprotectin in distinguishing severe disease, and a combined model significantly improved accuracy (P < 0.0001). Conclusion: Our findings demonstrate a connection between microbiome composition and the risk of severe disease development, which is partly independent of inflammation levels at the time of sampling. Microbiome-informed subgrouping could lead to more personalized treatment strategies. Further validation is needed to determine the clinical utility of these clusters.

  PublisherOA PDFDOI

• Resilient Path Tracking of Autonomous Driving under Few-shot Action Space Attacks

  ACM Transactions on Cyber-Physical Systems · 2025-11-21

  articleSenior author

  Modern autonomous vehicles face growing cybersecurity risks, especially from action space attacks that directly target vehicle actuators. This article systematically evaluates the resilience of three representative Autonomous Driving (AD) architectures, including modular, end-to-end, and feature-fused agents, against few-shot action space attacks crafted via deep reinforcement learning under a black-box setting. The adversary perturbs the vehicle’s lateral control only during safety-critical moments, using either a camera or an inertial measurement unit. Our results reveal distinct vulnerabilities and behavioral patterns across AD architectures, which underscore the necessity for adaptive and robust defense strategies. However, existing adversarial training defense methods show limitations of overfitting and reliance on attack knowledge. To address these limitations, we propose a learning-based Path Correction System (PCS) that integrates traditional feedback control with an adversarially trained correction loop. The correction loop is selectively activated by a kinematic model-based attack detector to counteract abnormal control deviations. Evaluation experiments show that PCS reduces path-tracking deviation by 78% when the system is under attack.

  PublisherDOI

• Page Migration for Hardware Memory Disaggregation Across a Network

  2025-06-08 · 1 citations

  articleOpen accessSenior author
  PublisherDOI

• Assessment of a Deep Learning Model Trained on Permanent Pathology for the Classification of Squamous Cell Carcinoma in Mohs Frozen Sections: Lessons Learned

  Dermatologic Surgery · 2025-07-09 · 2 citations

  article

  BACKGROUND: There is a scarcity of artificial intelligence models trained on frozen pathology. One way to expand the clinical utility of models trained on permanent pathology is by applying them to frozen sections and fine-tune based on weaknesses. OBJECTIVE: To qualitatively evaluate a deep learning model trained on permanent pathology to classify squamous cell carcinoma on Mohs surgery frozen sections to learn model shortcomings and inform retraining and fine-tuning. MATERIALS AND METHODS: The authors trained a model for classification of tumor on 746 skin biopsy slides and tested it on 15 Mohs surgery frozen sections. The authors estimated performance metrics and compared the regions of interest generated by the model with the original H&E slides. RESULTS: The model achieved an AUC-ROC of 0.985 and 0.796 for tumor classification in permanent pathology and in frozen sections, respectively. Regions of interest for frozen sections with scarce tumor areas were inaccurate, focusing on normal tissue for slides classified as false negative, or highlighting structures different from tumor (e.g., inflammation, muscle, and nerves) for slides classified as true positive. CONCLUSION: Deep anatomical structures more commonly present in Mohs frozen pathology might represent data out-of-distribution for models trained on permanent pathology, potentially leading to inadequate model outputs.

  PublisherDOI

• Hierarchical Autoscaling for Large Language Model Serving with Chiron

  ArXiv.org · 2025-01-14

  preprintOpen accessSenior author

  Large language model (LLM) serving is becoming an increasingly important workload for cloud providers. Based on performance SLO requirements, LLM inference requests can be divided into (a) interactive requests that have tight SLOs in the order of seconds, and (b) batch requests that have relaxed SLO in the order of minutes to hours. These SLOs can degrade based on the arrival rates, multiplexing, and configuration parameters, thus necessitating the use of resource autoscaling on serving instances and their batch sizes. However, previous autoscalers for LLM serving do not consider request SLOs leading to unnecessary scaling and resource under-utilization. To address these limitations, we introduce Chiron, an autoscaler that uses the idea of hierarchical backpressure estimated using queue size, utilization, and SLOs. Our experiments show that Chiron achieves up to 90% higher SLO attainment and improves GPU efficiency by up to 70% compared to existing solutions.

  PublisherOA PDFDOI

• PRAXIS: Integrating Program Analysis with Observability for Root-Cause Analysis

  arXiv (Cornell University) · 2025-12-26

  preprintOpen accessSenior author

  Unresolved production cloud incidents cost an average of over $2M per hour. This paper introduces PRAXIS, an orchestrator that manages and deploys an agentic workflow for diagnosing code- and configuration-caused cloud incidents. PRAXIS employs an LLM-driven structured traversal over two types of graph: (1) a service dependency graph (SDG) that captures microservice-level dependencies; and (2) a hammock-block program dependence graph (PDG) that captures code-level dependencies for each microservice. Compared to state-of-the-art ReAct baselines, PRAXIS improves RCA accuracy by up to 6.3x while reducing token consumption by 5.3x. PRAXIS is demonstrated on a set of 30 comprehensive real-world incidents that is being compiled into an RCA benchmark.

  PublisherDOI

Recent grants

• CPS: Breakthrough:Towards Resiliency in Cyber-physical Systems for Robot-assisted Surgery

  NSF · $500k · 2016–2019

• PPoSS: Planning: Inflight Analytics to Control Large-Scale Heterogeneous Systems

  NSF · $250k · 2020–2024

• CI: NEW: Collaborative Research: Computer System Failure Data Repository to Enable Data-Driven Dependability

  NSF · $371k · 2015–2019

• I/UCRC: Phase I: Computing and Genomics-An Essential Partnership for Biology Breakthroughs (CCBGM)

  NSF · $766k · 2016–2024

Frequent coauthors

• Zbigniew Kalbarczyk

  197 shared

• Saurabh Jha

  34 shared

• Subho S. Banerjee

  University of Illinois Urbana-Champaign

  28 shared

• Krishnakant Saboo

  26 shared

• Gregory A. Worrell

  WinnMed

  21 shared

• Arjun P. Athreya

  Mayo Clinic

  20 shared

• Yogatheesan Varatharajah

  University of Minnesota

  20 shared

• Phuong Cao

  Bucknell University

  17 shared

Education

• Ph.D., Computer Science

  University of Illinois at Urbana-Champaign

  2000

• M.S., Computer Science

  University of Illinois at Urbana-Champaign

  1996

• B.S., Electrical and Electronics Engineering

  University of Madras

  1993

Awards & honors

• IEEE Emanuel R. Piore Award
• Outstanding Contributions award by the Association of Comput…
• Fellow of the American Association for the Advancement of Sc…
• Fellow of the Institute of Electrical and Electronics Engine…
• Fellow of the Association for Computing Machinery (ACM)