About

Abhishek Chandra is a professor in the Department of Computer Science & Engineering at the University of Minnesota. His research interests are in the areas of operating systems and distributed systems, with a focus on resource management and performance in large-scale distributed systems. His work aims to achieve reliability, scalability, and manageability, particularly in data-intensive computing applications within cloud and edge computing platforms, especially for data generated near end-users such as mobile and IoT data. He joined the department in 2004 as an assistant professor and is affiliated with the Distributed Computing Systems Group Lab at the University. Chandra holds a Ph.D. and M.S. in computer science from the University of Massachusetts Amherst, and a B.Tech. in Computer Science and Engineering from the Indian Institute of Technology Kanpur. His professional background includes receiving the NSF CAREER Award in 2007 and the IBM Faculty Award in 2011. He is a lifetime member of ACM and a member of IEEE. His contributions include research on resource management in distributed systems, with recent recognition for best papers and posters at prominent conferences.

Research topics

• Computer Science
• Operating system
• Distributed computing
• Artificial Intelligence
• Machine Learning
• Data Mining
• Real-time computing
• Computer network
• Data science
• Business
• Embedded system

Selected publications

• Prospective study of bacteriological profile of indoor patient of respiratory diseases at tertiary care centre

  International Journal of Research in Medical Sciences · 2026-03-30

  articleOpen access1st authorCorresponding

  Background: Respiratory tract infections (RTIs) are a major cause of morbidity among hospitalized patients and are frequently associated with bacterial pathogens and antimicrobial resistance. Knowledge of the local bacteriological profile and antibiotic susceptibility pattern is essential for guiding appropriate empirical therapy and improving clinical outcomes. Methods: This prospective observational study was conducted over a period of one year from 20 January 2025 to 30 December 2025 at Rajarshi Dashrath Autonomous State Medical College, Ayodhya, Uttar Pradesh. A total of 102 indoor patients admitted with respiratory diseases were included. Respiratory samples, predominantly sputum, were collected and processed using standard microbiological techniques. Bacterial isolates were identified, and antibiotic susceptibility testing was performed. Data were analysed using the descriptive statistics and expressed as frequencies and the percentages. Results: Among the 102 patients studied, males constituted the majority. Gram-negative organisms were predominantly isolated. E. coli was the most common organism, followed by K. pneumoniae and P. aeruginosa. Most isolates showed higher sensitivity to broad-spectrum antibiotics such as carbapenems, aminoglycosides, and higher-generation cephalosporins, while resistance to commonly used first-line antibiotics was observed in several cases. Conclusions: Gram-negative bacteria were the predominant pathogens among indoor patients with respiratory diseases. Regular monitoring of bacteriological patterns and antibiotic susceptibility is essential to guide rational antibiotic therapy and to curb the growing problem of antimicrobial resistance in tertiary care settings.

  PublisherOA PDFDOI

• SneakPeek: Data-Aware Model Selection and Scheduling for Inference Serving on the Edge

  ArXiv.org · 2025-05-10

  preprintOpen accessSenior author

  Modern applications increasingly rely on inference serving systems to provide low-latency insights with a diverse set of machine learning models. Existing systems often utilize resource elasticity to scale with demand. However, many applications cannot rely on hardware scaling when deployed at the edge or other resource-constrained environments. In this work, we propose a model selection and scheduling algorithm that implements accuracy scaling to increase efficiency for these more constrained deployments. We show that existing schedulers that make decisions using profiled model accuracy are biased toward the label distribution present in the test dataset. To address this problem, we propose using ML models -- which we call SneakPeek models -- to dynamically adjust estimates of model accuracy, based on the underlying data. Furthermore, we greedily incorporate inference batching into scheduling decisions to improve throughput and avoid the overhead of swapping models in and out of GPU memory. Our approach employs a new notion of request priority, which navigates the trade-off between attaining high accuracy and satisfying deadlines. Using data and models from three real-world applications, we show that our proposed approaches result in higher-utility schedules and higher accuracy inferences in these hardware-constrained environments.

  PublisherOA PDFDOI

• SneakPeek: Data-Aware Model Selection and Scheduling for Inference Serving on the Edge

  2025-11-19

  articleSenior author

  Modern applications increasingly rely on inference serving systems to provide low-latency insights with a diverse set of machine learning models. Existing systems often utilize resource elasticity to scale with demand. However, many applications cannot rely on hardware scaling when deployed at the edge or other resource-constrained environments. In this work, we propose a model selection and scheduling algorithm that implements accuracy scaling to increase efficiency for these more constrained deployments. We show that existing schedulers that make decisions using profiled model accuracy are biased toward the label distribution present in the test dataset. To address this problem, we propose using ML models-which we call SneakPeek models- to dynamically adjust estimates of model accuracy, based on the underlying data. Furthermore, we greedily incorporate inference batching into scheduling decisions to improve throughput and avoid the overhead of swapping models in and out of GPU memory. Our approach employs a new notion of request priority, which navigates the trade-off between attaining high accuracy and satisfying deadlines. Using data and models from three real-world applications, we show that our proposed approaches result in higher-utility schedules and higher accuracy inferences in these hardware-constrained environments.

  PublisherDOI

• ASTRA: Association, Spatial proximity and Temporal Relevance based Adaptive prefetching for Edge AR

  2025-09-23

  article

  Mobile Augmented Reality (MAR) applications face performance challenges due to their high computational demands and need for low-latency responses. Traditional approaches like on-device storage or reactive data fetching from the cloud often result in limited augmented reality (AR) experiences. Edge caching, which caches AR objects closer to the user, provides a promising solution. However, existing edge caching approaches do not consider AR-specific features such as AR object sizes, user interactions, user’s field of view and physical location in a coherent manner. This paper investigates how to further optimize edge caching by employing AR-aware prefetching techniques. We present ASTRA, a prefetching framework tailored for mobile augmented reality edge caches. It integrates object associations derived from user interaction patterns with spatial awareness based on the user’s physical location and field of view. This approach employs an association factor per object that considers the recency of object co-access; and a lazy fetching strategy that prioritizes prefetching only when the user is in close proximity to the virtual objects. Furthermore, ASTRA incorporates an adaptive tuning algorithm for minimum support in association rule generation to minimize the computation overhead, making it a distinct and effective solution for enhancing user experience in AR applications by ensuring timely virtual object availability.Through extensive evaluation using both synthetic and real-world workloads, we demonstrate that ASTRA significantly improves cache hit rates compared to current prefetching algorithms, achieving gains in hit rate of upto 35% and end-to-end latency by upto 14%. Further, we demonstrate that the adaptive tuning algorithm that automatically tunes minimum support further improves the hit rate of ASTRA by 10%. Our findings demonstrate the potential of ASTRA to substantially enhance the user experience in MAR applications by ensuring the timely availability of virtual objects.

  PublisherDOI

• SPAARC: Spatial Proximity and Association based prefetching for Augmented Reality in edge Cache

  ArXiv.org · 2025-02-21

  preprintOpen access

  Mobile Augmented Reality (MAR) applications face performance challenges due to their high computational demands and need for low-latency responses. Traditional approaches like on-device storage or reactive data fetching from the cloud often result in limited AR experiences or unacceptable lag. Edge caching, which caches AR objects closer to the user, provides a promising solution. However, existing edge caching approaches do not consider AR-specific features such as AR object sizes, user interactions, and physical location. This paper investigates how to further optimize edge caching by employing AR-aware prefetching techniques. We present SPAARC, a Spatial Proximity and Association-based Prefetching policy specifically designed for MAR Caches. SPAARC intelligently prioritizes the caching of virtual objects based on their association with other similar objects and the user's proximity to them. It also considers the recency of associations and uses a lazy fetching strategy to efficiently manage edge resources and maximize Quality of Experience (QoE). Through extensive evaluation using both synthetic and real-world workloads, we demonstrate that SPAARC significantly improves cache hit rates compared to standard caching algorithms, achieving gains ranging from 3% to 40% while reducing the need for on-demand data retrieval from the cloud. Further, we present an adaptive tuning algorithm that automatically tunes SPAARC parameters to achieve optimal performance. Our findings demonstrate the potential of SPAARC to substantially enhance the user experience in MAR applications by ensuring the timely availability of virtual objects.

  PublisherOA PDFDOI

• A Comprehensive Study on Factors Leading to Nonadherence to Inhaler Therapy Among Chronic Obstructive Pulmonary Disease Patients in a Tertiary Care Setting

  Indian Journal of Health Sciences and Biomedical Research (KLEU) · 2025-01-30

  articleOpen access

  INTRODUCTION: Medication adherence is crucial for managing respiratory diseases. Understanding the factors influencing adherence among patients with respiratory illnesses is essential for improving outcomes and optimizing treatment. OBJECTIVE: This study explored the multifaceted factors influencing medication adherence among patients with respiratory illnesses, focusing on patient characteristics, clinical factors, beliefs about medications, and reasons for irregular medication use. METHODOLOGY: A cross-sectional design assessed medication adherence and beliefs in chronic obstructive pulmonary disease (COPD) patients through questionnaires. Patient demographics and clinical characteristics were analyzed to understand their impact on adherence. STATISTICAL ANALYSIS USED: Medication adherence was measured using the Adherence to Refills and Medications Scale (ARMS). Descriptive statistics summarized patient demographics and clinical characteristics. Correlations between various factors and adherence levels were determined. RESULTS: The study identified a middle-aged population with an average age of 48.5 years and a high smoking prevalence, aligning with COPD risk factors. Medication adherence scores indicated moderate levels (average ARMS score: 62.7), with significant individual variation. Financial burden was the primary barrier to adherence, with nearly 28% reporting cost-related nonadherence, particularly among lower socioeconomic status (SES) patients. Social stigma influenced adherence, especially in urban areas (19.8% reported stigma). Positive beliefs about medications correlated with better adherence. A confirmed COPD and asthma diagnosis and using specific inhaler devices (dry powder inhaler, PMDI with spacer) were linked to improved adherence. CONCLUSION: Significant factors impacting medication adherence in COPD include financial constraints, social stigma, medication beliefs, and inhaler device type. Addressing these through cost-reduction strategies, stigma interventions, and educational programs can improve COPD management and outcomes, especially for older adults and lower SES populations.

  PublisherOA PDFDOI

• Leveraging Multi-Modal Data for Efficient Edge Inference Serving

  2024-05-06

  articleSenior author

  Real-time analytics over data streams is often performed on edge devices, which offer privacy guarantees and lower-latency responses compared to centralized processing in the cloud. Data streams originating from sensors, mobile phones, or IoT devices are diverse and span multiple modalities, including RGB videos from cameras, time series data from wearable sensors, and audio signals. Previous research has focused on optimizing the individual analytical tasks associated with each stream, with a special emphasis on deep learning, which is computationally intensive and may be used to analyze video streams, among other things. While advances in deep learning have significantly improved inference accuracy (e.g. for computer vision tasks), state-of-the-art models are not well-suited for edge computing environments. Novel approaches are required to substantially reduce the computational burden, since edge systems are heterogeneous and typically have fewer GPU resources available for inference with deep learning models. We show that leveraging data from multiple modalities can complement or sometimes even replace resource-intensive inference, while maintaining or enhancing accuracy. We present DAISY: a Data-Aware Inference Serving sYstem which leverages multi-modal data to increase inference accuracy by dynamically selecting an appropriate model for each request. We thoroughly evaluate the proposed approach using state-of-the-art models and real-world data, which shows an increase in SLO attainment up to 60%, with a corresponding increase in inference accuracy of 5%.

  PublisherDOI

• A Biased Estimator for MinMax Sampling and Distributed Aggregation

  arXiv (Cornell University) · 2024-04-26

  preprintOpen accessSenior author

  MinMax sampling is a technique for downsampling a real-valued vector which minimizes the maximum variance over all vector components. This approach is useful for reducing the amount of data that must be sent over a constrained network link (e.g. in the wide-area). MinMax can provide unbiased estimates of the vector elements, along with unbiased estimates of aggregates when vectors are combined from multiple locations. In this work, we propose a biased MinMax estimation scheme, B-MinMax, which trades an increase in estimator bias for a reduction in variance. We prove that when no aggregation is performed, B-MinMax obtains a strictly lower MSE compared to the unbiased MinMax estimator. When aggregation is required, B-MinMax is preferable when sample sizes are small or the number of aggregated vectors is limited. Our experiments show that this approach can substantially reduce the MSE for MinMax sampling in many practical settings.

  PublisherOA PDFDOI

• A review on progresses in solar still technology with phase change material

  International Journal of Mechanical and Thermal Engineering · 2024-01-01 · 2 citations

  reviewOpen access1st authorCorresponding

  The utilization of solar desalination processes using solar stills to maintain the purity of water. The purpose of this process is to remove impurities and salt from water to make it suitable for consumption. The study involves analyzing various factors that influence the conditions of the solar desalination process at different stages. The article reviews the application of highly efficient materials and experimental endeavors in the realm of solar stills. One notable technique involves using Paraffin wax in combination with different Phase Change Materials (PCMs). This combination leads to increased productivity and thermal conductivity, ultimately enhancing the output of the solar still. In a modified solar still, the incorporation of wicks resulted in amplified distillate output and an overall improvement in productivity. Another enhancement strategy discussed is the utilization of nano-particles. The incorporation of these nano-particles led to improved daily efficiency and thermal conductivity. This, in turn, raised the thermal efficiency of the solar still and augmented the yield of fresh water. The article also underscores the efficacy of collective approaches involving Phase Change Materials, reflectors, and nano-coating paint mixed with nano-particles. These combined materials resulted in a notable increase in thermal efficiency and fresh-water yield. Among the various Phase Change Materials explored, the study concludes that paraffin wax exhibits the highest output and productivity compared to other options. In essence, the article highlights the advancements and innovations in solar desalination using solar stills.

  PublisherOA PDFDOI

• SIOG2024-3-OA-029 Preliminary findings from the communication on aging and cancer health (COACH 2.0) Implementation project: what it takes to implement the geriatric assessment (GA) In the real world

  Journal of Geriatric Oncology · 2024-10-01

  article
  PublisherDOI

Recent grants

• CSR: Small: Collaborative Research: Dispersed Real-time Data Analytics

  NSF · $258k · 2017–2022

• CAREER: Self-Managing Resource Allocation in Unsupervised Distributed Systems

  NSF · $416k · 2007–2012

• III: Small: MESH: A Hypergraph Analysis Engine for Understanding Large-Scale Social Networks

  NSF · $516k · 2014–2019

• CSR: Small: Location, location, location (L3): Support for Geo-Centric Applications

  NSF · $516k · 2016–2020

Frequent coauthors

• Jon Weissman

  University of Minnesota

  95 shared

• Albert Jonathan

  Tarumanagara University

  25 shared

• Ramesh K. Sitaraman

  University of Massachusetts Amherst

  25 shared

• Kwangsung Oh

  University of Nebraska at Omaha

  20 shared

• Benjamin Heintz

  University of Minnesota

  20 shared

• Prashant Shenoy

  15 shared

• Dhruv Kumar

  University of Minnesota System

  12 shared

• Michael Cardosa

  10 shared

Labs

• Distributed Computing Systems LabPI

Awards & honors

• IBM Faculty Award (2011)
• National Science Foundation Faculty Early Career Development…