About

Volodymyr Kindratenko is an Assistant Director at the National Center for Supercomputing Applications (NCSA) at the University of Illinois at Urbana-Champaign, where he also serves as the Director for the Center for Artificial Intelligence Innovation (CAII). He holds appointments as an Adjunct Associate Professor in the Departments of Electrical and Computer Engineering and as a Research Associate Professor in the Department of Computer Science. His academic background includes a D.Sc. in analytical chemistry from the University of Antwerp and an M.Sc. in mathematics and informatics from Volodymyr Vynnychenko Central Ukrainian State University. Dr. Kindratenko's research interests encompass high-performance computing, special-purpose computing architectures, cloud computing, and machine learning systems and applications. He has led research efforts in developing next-generation HPC systems based on computational accelerators and designing scientific and AI-driven applications for such systems. His work has been funded by NSF, NASA, ONR, DOE, and industry, and he has published extensively, including over 140 papers and holding five US patents. He also serves as an associate editor of the International Journal of Reconfigurable Computing. His contributions include editing books on GPU numerical computations and accelerator technologies for geographic information science, reflecting his expertise in high-performance and specialized computing architectures.

Research topics

• Artificial Intelligence
• Computer Science
• Machine Learning
• Data science
• Political Science
• Distributed computing
• Computer architecture
• Embedded system
• Engineering ethics
• World Wide Web
• Parallel computing
• Operating system
• Computational science
• Knowledge management
• Database
• Engineering

Selected publications

• DeltaAI: A National Resource for AI/ML Research

  2025-07-18 · 1 citations

  articleOpen access

  DeltaAI is a new NSF funded resource supporting researchers nationwide via the ACCESS and NAIRR Pilot programs. DeltaAI builds upon the success of Delta and supplants Delta as the most powerful GPU resource available via the ACCESS program. DeltaAI leverages the Delta environment sharing storage and infrastructure resources while offering a more scalable platform capable of scaling jobs up to the full node count of the system. DeltaAI offers the latest NVIDIA H100 GPUs as part of the innovative Grace Hopper architecture. This paper describes the full architecture of DeltaAI and our experiences benchmarking applications and supporting research teams during the acceptance process for DeltaAI.

  PublisherDOI

• Frameworks for Large Language Model Serving in HPC Environments

  2025-11-07 · 1 citations

  articleOpen accessSenior author

  We introduce open-source frameworks for deploying and running large language models (LLMs) within high-performance computing (HPC) environments. One such framework, AI-Flux, targets high-throughput batch inference, enabling users to submit LLM requests in an OpenAI-compatible format as traditional HPC jobs. Another framework is based on Ray Serve and it provides dynamic, on-demand allocation of HPC resources for interactive LLM serving via APIs, supporting applications such as chatbots and AI agents. The third framework, Illinois Chat, is a production-grade, always-on platform for real-time interaction, that relies on a dedicated GPU server for model inference. These frameworks are designed to abstract away underlying computer system complexities, allowing researchers to request and utilize GPU resources for model inference without manual environment setup. We describe these systems and report LLM-specific performance metrics. Results demonstrate that the proposed frameworks enable scalable and resource-efficient LLM serving across both batch and interactive workloads in support of a diverse user needs.

  PublisherDOI

• Diamond: Harnessing GPU Resources for Scientific Deep Learning

  2025-09-15

  article

  Modern research computing cyberinfrastructure, such as ACCESS-CI and NAIRR Pilot, offers GPU resources across geographically distributed clusters to accommodate the increasing needs of scientific deep learning (DL) workloads. Even for high-performance computing (HPC) experts, configuring environments and managing DL workloads across supercomputers remain significant barriers. To address these obstacles, we present Diamond, an open-source platform to simplify and streamline the DL lifecycle on HPC. Diamond provides an intuitive graphical interface that abstracts system-level complexity, enabling users to develop, debug, and deploy DL models with minimal overhead. We identify several challenges in building such a platform, including portability, security, and usability, and propose effective architectural solutions to each. Notably, Diamond enables users to share and reuse DL workload environments across systems and collaborators, reducing redundant setup efforts. Experimental results demonstrate that Diamond reduces the time to first successful deployment by an average of 68%, compared to manual configuration with command lines. The Diamond service is available at https://diamondhpc.ai.

  PublisherDOI

• How Students Use Generative AI: Insights from Conversation Log Analysis

  2025-11-02

  article

  This paper aims to provide insights into how students interact with Generative Artificial Intelligence (GenAI) and identify patterns in their usage of Illinois Chat, a GenAI educational tool developed by the University of Illinois Urbana Champaign. With advancements in technology and machine learning, GenAI has become increasingly prevalent across various domains, including education. It is evident that students frequently use GenAI tools like ChatGPT in their academic lives; however, there are limited studies analyzing how they integrate these tools into their daily learning practices. To address this gap, we propose a study that examines how students interact with a GenAI tool across different engineering disciplines. Specifically, we aim to identify the common categories of questions students seek help with, discern patterns in their input prompts, and examine potential instances of academic dishonesty involving the GenAI tool. Using both inductive and deductive thematic analysis, we analyzed conversation log data to uncover emerging themes and validate prior findings. Our analysis revealed distinct patterns in how students use the tool to understand concepts, their behavioral trends during interactions, and signs that may indicate academic dishonesty. This study provides insights into how students engage with GenAI and highlights areas for improvement to enhance the educational effectiveness of these tools. Additionally, we outline potential indicators of academic dishonesty, which could help mitigate learning loss in the AI-driven educational landscape.

  PublisherDOI

• The role of experience with visual and linguistic factors on skipping behavior during reading in teen and adult readers

  Journal of Vision · 2025-07-15

  articleOpen access

  During reading skilled readers skip and never fixate on between 20-30% of words. In controlled experiments, a combination of low-level visual factors such as word length and higher-level linguistic factors such as word frequency and lexical predictability affects skilled readers’ skipping: shorter and lexically more predictable words are skipped more often. Recently, syntactic predictability (expectations about the upcoming word’s part of speech and not the exact word) has been proposed as an additional source of linguistic predictability that affects skipping. To understand how skipping behavior emerges with experience, we investigated skipping behavior in both adults and adolescents, who are skilled readers but not yet adult-like in their behavior. The present study examined how visual (word length) and linguistic (word frequency, lexical and syntactic predictability) knowledge during reading affects skipping in skilled and adolescent readers. 113 college students and 52 adolescents (14-17 yoa) read 55 passages from PROVO corpus (Luke & Christianson, 2018) along with vocabulary (Shipley, 1941) and reading comprehension (shortened 10 minute version of Nelson&Denny, 1980) tests, combined into one reading experience composite score. Logistic mixed-effects regression examined the effects of reading experience, word length, word frequency, lexical and syntactic predictability controlling for the position in the sentence. Our results replicated prior findings from experimental and naturalistic work: readers skip predictable short words more often than predictable long words. Further, two novel findings emerged. First, word length interacted with reading experience such that better readers skip more longer words than poorer readers. Second, word length interacted with syntactic predictability such that longer, syntactically predictable words are skipped more than less predictable words. This work highlights the role of readers’ experience with linguistic knowledge and word's syntactic predictability in eye movements. We show tighter than previously assumed coupling between readers’ eyes movements and higher-level linguistic predictability beyond specific lexical items.

  PublisherDOI

• Cost-Aware Federated Learning on the Cloud

  2025-09-15 · 1 citations

  article

  We introduce FedCostAware, a cost-aware scheduling algorithm designed to optimize synchronous federated learning (FL) on cloud spot instances, which addresses the challenges of training on spot instances and different client budgets by employing intelligent management of the lifecycle of spot instances. This approach minimizes idle resource time and overall expenses. Experiments on real-world medical datasets demonstrate that FedCostAware significantly reduces cloud computing costs compared to conventional spot and on-demand schemes, enhancing the accessibility and affordability of FL.

  PublisherDOI

• Evidential Deep Learning for Uncertainty Quantification and Out-of-Distribution Detection in Jet Identification using Deep Neural Networks

  arXiv (Cornell University) · 2025-01-10 · 2 citations

  preprintOpen access

  Current methods commonly used for uncertainty quantification (UQ) in deep learning (DL) models utilize Bayesian methods which are computationally expensive and time-consuming. In this paper, we provide a detailed study of UQ based on evidential deep learning (EDL) for deep neural network models designed to identify jets in high energy proton-proton collisions at the Large Hadron Collider and explore its utility in anomaly detection. EDL is a DL approach that treats learning as an evidence acquisition process designed to provide confidence (or epistemic uncertainty) about test data. Using publicly available datasets for jet classification benchmarking, we explore hyperparameter optimizations for EDL applied to the challenge of UQ for jet identification. We also investigate how the uncertainty is distributed for each jet class, how this method can be implemented for the detection of anomalies, how the uncertainty compares with Bayesian ensemble methods, and how the uncertainty maps onto latent spaces for the models. Our studies uncover some pitfalls of EDL applied to anomaly detection and a more effective way to quantify uncertainty from EDL as compared with the foundational EDL setup. These studies illustrate a methodological approach to interpreting EDL in jet classification models, providing new insights on how EDL quantifies uncertainty and detects out-of-distribution data which may lead to improved EDL methods for DL models applied to classification tasks.

  PublisherOA PDFDOI

• Highly Accelerated T1 and T2 Mapping in Water-Unsuppressed MRSI Using Generalized Series Modeling and Deep Learning

  Proceedings on CD-ROM - International Society for Magnetic Resonance in Medicine. Scientific Meeting and Exhibition/Proceedings of the International Society for Magnetic Resonance in Medicine, Scientific Meeting and Exhibition · 2025-09-16

  article

  Motivation: MRSI and quantitative parametric mapping provide complementary tissue information, but separate acquisitions result in long scan times. Goal(s): To achieve 1-mm whole-brain T1/T2 mapping in water-unsuppressed MRSI for rapid simultaneous metabolic and parametric mapping. Approach: In data acquisition, we further accelerated the SPICE sequence for T1/T2 mapping using shorter TR, extended readout, and sparser sampling. In image reconstruction, we proposed a novel model-based method to incorporate spatiotemporal priors from metabolic imaging and training data for T1/T2 reconstruction from highly sparse data. Results: Phantom and in vivo experiments showed the proposed method was accurate and reproducible. Impact: The proposed method provides a powerful multimodal imaging capability that provides tissue structural and biochemical biomarkers at the same time. This new imaging technology may enable better tissue characterization desired for various research and clinical applications.

  PublisherDOI

• ORBIT: Cost-Effective Dataset Curation for Large Language Model Domain Adaptation with an Astronomy Case Study

  2025-01-01

  articleOpen accessSenior author

  Recent advances in language modeling demonstrate the need for high-quality domain-specific training data, especially for tasks that require specialized knowledge.General-purpose models, while versatile, often lack the depth needed for expert-level tasks because of limited domain-specific information.Domain adaptation training can enhance these models, but it demands substantial, high-quality data.To address this, we propose ORBIT, a costefficient methodology for curating massive, high-quality domain-specific datasets from noisy web sources, tailored for training specialist large language models.Using astronomy as a primary case study, we refined the 1.3T-token FineWeb-Edu dataset into a highquality, 10B-token subset focused on astronomy.Fine-tuning LLAMA-3-8B on a 1Btoken astronomy subset improved performance on the MMLU astronomy benchmark from 69% to 76% and achieved top results on As-troBench, an astronomy-specific benchmark.Moreover, our model (Orbit-LLaMA) outperformed LLAMA-3-8B-BASE, with GPT-4o evaluations preferring it in 73% of cases across 1000 astronomy-specific questions.Additionally, we validated ORBIT's generalizability by applying it to law and medicine, achieving a significant improvement of data quality compared to an unfiltered baseline.We open-source the ORBIT methodology, including the curated datasets, the codebase, and the resulting model at https://github.com/ModeEric/ORBIT-Llama.

  PublisherOA PDFDOI

• Hierarchical Diffusion Framework for Pseudo-Healthy Brain MRI Inpainting with Enhanced 3D Consistency

  ArXiv.org · 2025-07-23

  preprintOpen accessSenior author

  Pseudo-healthy image inpainting is an essential preprocessing step for analyzing pathological brain MRI scans. Most current inpainting methods favor slice-wise 2D models for their high in-plane fidelity, but their independence across slices produces discontinuities in the volume. Fully 3D models alleviate this issue, but their high model capacity demands extensive training data for reliable, high-fidelity synthesis -- often impractical in medical settings. We address these limitations with a hierarchical diffusion framework by replacing direct 3D modeling with two perpendicular coarse-to-fine 2D stages. An axial diffusion model first yields a coarse, globally consistent inpainting; a coronal diffusion model then refines anatomical details. By combining perpendicular spatial views with adaptive resampling, our method balances data efficiency and volumetric consistency. Our experiments show our approach outperforms state-of-the-art baselines in both realism and volumetric consistency, making it a promising solution for pseudo-healthy image inpainting. Code is available at https://github.com/dou0000/3dMRI-Consistent-Inpaint.

  PublisherOA PDFDOI

Recent grants

• Collaborative Research: Frameworks: Machine learning and FPGA computing for real-time applications in big-data physics experiments

  NSF · $651k · 2019–2023

• REU Site: The future of discovery: training students to build and apply open source machine learning models and tools

  NSF · $405k · 2021–2025

• SGER: Investigating Application Analysis and Design Methodologies for Computational Accelerators

  NSF · $166k · 2008–2009

Frequent coauthors

• Guochun Shi

  Oeschger Centre for Climate Change Research

  48 shared

• E. A. Huerta

  35 shared

• Steven Gottlieb

  Indiana University Bloomington

  26 shared

• Shirui Luo

  22 shared

• Daniel S. Katz

  20 shared

• Xin Liu

  16 shared

• Aaron Torok

  National Institute of Science Education and Research

  14 shared

• M. Carrasco Kind

  Urbana University

  14 shared

Labs

• Siebel School of Computing and Data SciencePI

Education

• Ph.D., Computer Science

  University of Illinois at Urbana-Champaign

  2005

• M.S., Computer Science

  University of Illinois at Urbana-Champaign

  2001

• B.S., Computer Science

  University of Illinois at Urbana-Champaign

  1998

Awards & honors

• Celebration of Excellence (2021)
• Celebration of Excellence (2022)
• Celebration of Excellence (2023)
• Celebration of Excellence (2024)
• Celebration of Excellence (2025)