Research 

To explore the possibility of developing a complete software stack for a cluster of heterogeneous Noisy Intermediate-Scale Quantum (NISQ) machines. The main objective is to make contributions to the following areas:

• Realizing a cluster of heterogeneous NISQ machines as a quantum-computing platform with large-scale simulation and evaluation on a real platform.

• Developing a programming environment and user interface to provide a visual interface to understand quantum noise.

• Developing compilation techniques that can account for the heterogeneity of NISQ machines and temporal errors.

• Creating a runtime that enables fault-tolerance, resource management, and scheduling while considering real-time queuing time and noise conditions. A resource monitoring mechanism will also query the calibration information from all available quantum computers.

• Co-designing the stack with quantum machine learning and quantum chemistry applications.

• Utilizing the system calibration data from the multiple existing quantum machines and then applying fidelity degradation detection on each noise attribute to generate the fidelity degradation matrix. This matrix will be used to define multiple new evaluation metrics to compare the fidelity between the qubit topology of the quantum machines.

This effort is supported by NSF Awards CCF 2216898, CCF 2216519, CCF 2217021, and CCF 2216923.

_____________________________________________________________________________

To develop artificial intelligence, specifically machine learning algorithms, that can be trained quickly, with minimal errors, and without requiring significant human expertise. We will utilize emerging quantum computing (QC) algorithms that have the potential to rapidly train models and find better solutions in a short time. We will focus on utilizing QC in machine learning to demonstrate its effectiveness in two challenging applications:

• Detecting seizures on encephalography signals.

• Interpreting digital pathology images automatically.

The successful implementation of these applications will enable automated systems to approach human performance and increase the impact of this technology in the medical field, which will benefit countless people globally. We will use adiabatic quantum annealing (QA) to solve two significant computational challenges:

• Finding a global minimum.

• Sampling from complex probability distributions.

We will also demonstrate that QA-supported sampling in training can find better parameters than conventional parameter optimization approaches and overcome the deficiencies of current machine learning algorithms in challenging applications, such as seizure detection on encephalography signals and automatic interpretation of digital pathology images. Moreover, our goal is to demonstrate that QA can easily sample a wide range of configuration spaces (undirected probabilistic graphical models trained with a variety of application-relevant data) that make it difficult to find local valleys in the probability distribution. These findings will be applied to deep generative models for superior classification and pattern reconstruction accuracy. The ability of QA to reach challenging sample regions of the configuration space will benefit many machine-learning applications.

This effort is supported by NSF Award CCF 2211841.

___________________________________________________________________________

Rapid advances are occurring in the field of quantum computing; however, even with future hardware enhancements and noise reduction, standalone quantum devices will be insufficient to solve many large-scale, real-world problems. This limitation stems from a fundamental constraint in the number of available logical qubits. To overcome this, the development of hybrid quantum-classical systems, which integrate quantum computing with high-performance computing (HPC), is essential.

This project focuses on the application-aware co-design of hybrid algorithms to achieve a practical quantum advantage in optimization and machine learning tasks. Such tasks are prevalent in fields such as logistics, medical signal analysis, and materials science, all of which demand real-time and large-scale processing. A large-scale project is necessary to integrate expertise across quantum algorithms, high-performance computing systems, and domain-specific applications. 

This effort is supported by NSF Award CNS 2444042.

We will investigate a new and innovative family of unconventional resource strategies that leverage Bayesian Optimization (BO) principles to optimize data center resource management. We aim to capture the impact of resource allocation on application performance using learning models based on BO. It will partition shared resources and adjust hardware and software settings to maximize individual application performance and overall system utilization.

To ensure practicality and scalability, we will employ a pool of lightweight, approximately-accurate online learning models instead of a heavy, fully-accurate model. Moreover, our runtime framework will place and co-locate incoming applications with existing ones in an efficient, dynamic, and non-intrusive manner.

The ultimate goal is to improve the utilization of large-scale data centers and HPC systems. This improvement will lead to better cost savings and a lower carbon footprint, which will have a significant impact on the operations of modern data centers. These data centers serve our computational needs for a variety of workloads, including short-running latency-critical applications (e.g., machine learning inferences, web search queries, microservices) and long-running throughput-oriented workloads (e.g., scientific simulations).

This effort is supported by NSF Awards CNS 2124908 and CNS 2124897.

_____________________________________________________________________________

We aim to develop the mathematical foundations, theoretical limits, and efficient implementation of Elastic Interval Runtime Schedulers (EIRIS). Our effort will consist of three main areas:

• Establishing the mathematical foundations of EIRIS by using group theory concepts and operations to create an analysis oracle, calculate theoretical lower bounds, and refine them further.

• Developing resource-orchestration techniques for EIRIS using bin-packing algorithms, which will rely on both the actual job duration times and the remaining job duration times.

• Implementing an autonomic runtime for EIRIS in G-Hadoop, which will include distributed resource management systems, metadata servers, and reproducible artifact packages.

To demonstrate the effectiveness of these three areas of research, we will deploy and test them on large-scale computer systems.

This effort is supported by NSF Award CCF 2135439.

This research project is addressing the connection between social media promotion of tobacco and increased tobacco use among youth. The project aims to develop new artificial intelligence (AI) approaches to automatically detect tobacco promotion and understand its impact on young people.

We plan to create a system that uses multi-modality, privacy-preserved machine learning, and human-in-the-loop methodologies. This involves several key steps. A comprehensive pipeline will be established to gather a large, multimodal dataset of tobacco-related social media content. The project will then develop multimodal AI models using adaptive vision-language transformers and debiased learning to ensure equitable analysis. To protect data, a new privacy-preserving federated learning system will be used, incorporating differential privacy and secure multi-party computation. A fairness continual learning system will also be introduced to ensure ethical alignment as the AI adapts to new data.

The project's outcomes will include a portable AI-based software for the early detection of tobacco promotion threads on social media. All developed algorithms will be deployed as cloud services and released as open source to benefit public health research. This work is expected to provide new theoretical and practical approaches for analyzing the impact of social media on youth tobacco use and will help guide targeted interventions.

This effort is supported by NSF Award IIS 2501021.

_____________________________________________________________________________

Native bee species are in decline, a significant problem given their $3.5 billion annual contribution to U.S. agricultural pollination. To address this, we are developing AI technology for real-time, in-field identification and tracking of pollinator species. This system will combine different cutting-edge AI techniques to learn and adapt, becoming more accurate and user-friendly over time. The ultimate goal is to create easy-to-use software that provides scientists and conservationists with valuable insights into how environmental changes affect these crucial species. By tracking pollinator activity and assessing how forest landscape changes impact abundance and diversity in the southeastern and northeastern United States, this information can be integrated into harvest scheduling programs for forest companies. This will help them with conservation planning, a key component of sustainable forest certification, ultimately informing and strengthening strategies to protect pollinators and the ecosystems that depend on them.

This effort is supported by NSF Awards DEB 2529183 and DEB 2529184.

We aim to analyze and strategize counter unmanned aerial vehicle (UAV) systems. We will analyze various UAVs and make recommendations on their capabilities, scope, and penetrability.

This effort is supported by FAA and DHS Award A60.

Edge computing systems are designed to focus on specific domains, such as healthcare, vehicles, and factories, which improves the user experience. These systems can make timely and accurate intelligent decisions tailored to each domain's needs. The Intelligent Edge Computing Systems (iEDGE) REU site is an initiative that exposes undergraduate students to various domain-specific edge computing systems. Eight undergraduate students will work on iEDGE projects each year for ten weeks. The REU site provides an environment that focuses on peer mentoring, rapid integration of students into existing projects, and robust support systems, which include knowledge transfer from faculty members, postdoctoral researchers, and doctoral, master, and undergraduate research students already working on edge computing projects at Mississippi State University.

This effort is supported by NSF Award CNS 2348711.

To develop cyberinfrastruture training modules tailored to the construction management and safety (CMS) domain. The overarching goal of the project is to equip trainees and students with practical, industry-relevant skills. The training modules will encompass four key areas: integration of IoT, ML/DL, robotics, and cybersecurity within the CMS field, covering essential aspects such as hardware, software, networking, and understanding cyber-attacks and defenses.

This effort is supported by NSF Award OAC 2417396.

We aim to bring together researchers and administrators from the Established Program to Stimulate Competitive Research (EPSCoR) states and territories to identify how their institutions can better contribute to the fields of quantum computing and quantum information science and engineering.

This effort is supported by NSF Award OIA 2202377.

_____________________________________________________________________________

We provide support for individuals who wish to attend the Institute of Electrical and Electronics Engineers (IEEE) Cloud Summit. The summit offers a unique platform for researchers, developers, and practitioners to showcase their cutting-edge research and share their best practices. It also provides an opportunity to exchange ideas and views on current trends and technical challenges related to Cloud Computing, Fog Computing, the Internet of Things, and Edge Computing, as well as their applications for societal benefit.

This effort is supported by NSF Award CNS 2243579.