• Abdelrahman Abdelkader, Ph.D. / Electrical and Computer Engineering / Symposium Talk 

  • Project Summary: Modern power systems witness an increased penetration of distributed energy resources (DER). DER aggregators act as the point-of-contact to coordinate operation between DER and grid operator. This requires an aggregated control framework in addition to local DER controls. Unlike traditional power plants, DER aggregators lack the centralized communication and computation infrastructure to perform such tasks. A cloud-based framework can be utilized to provide the needed infrastructure and the flexibility to accommodate geographically diverse DER. Using AWS, this project develops a framework to perform the required coordinated control tasks. Communication needs are met by registering DER controllers as IoT things in the AWS IoT core. The required computations are performed using an AWS Lambda function to respond to operator demand and enforce system-wide constraints.

   

• Ramin Esmzad, Ph.D. / Mechanical Engineering / Symposium Talk 

  • Project Summary: This project focuses on building and deploying an online Model Predictive Control (MPC) simulation platform using AWS cloud resources. The platform features a Django backend with Celery, RabbitMQ, and Redis to support asynchronous task processing and real-time communication, paired with a Next.js frontend for an interactive user experience. It enables users to define custom system models and run real-time MPC simulations directly in the cloud. The system is deployed using Amazon EC2 for scalable compute, Amazon RDS for relational data management, and AWS CloudWatch for monitoring and logging application performance. IAM is used to manage secure access to resources. The platform helps researchers conduct better and faster research by leveraging robust and scalable cloud infrastructure.

   

• Raaghav Ravishankar, Ph.D. Candidate / Computer Science and Engineering / Symposium Talk 

  • Project summary: My project is an asynchronously partitioned and load balanced distributed linked list. The goal of the project is to measure the overhead of distribution by observing the variation in throughput and latency of client operations to the linked list by experimental simulations. These experiments are performed in a client-server model, with AWS EC2 instances as the servers of the linked list, and computing clusters at MSU as the many clients that send linked list operation requests (workload) to the EC2 instances (permitted through Amazon VPC). The experiment is performed by varying many parameters to the distributed list configuration to push the EC2 instances to maximum CPU utilization (monitored via AWS Cloud Watch). Measurements gathered at this point provides us metrics during optimal performance.

   

• Leah Terrian, Ph.D. Candidate / Biomedical Engineering / Symposium Talk 

  • Project Summary: During my PhD, I have encountered many instances where people who are generating data cannot easily access the analytical tools needed to process it. I aimed to resolve that by using AWS Lightsail to deploy a web-application that would make data analysis, specifically single-cell RNA-sequencing data analysis, more accessible to the people who need it most.

   

• Hardik Arora, MSc / Computer Science and Engineering / Symposium Talk

  • Project Summary: This project is designed to predict stock market prices using machine learning models deployed on AWS. It integrates multiple AWS services to process historical and real-time stock data, train machine learning models, and generate predictions that can be accessed through an API.

   

• Aryan Gondkar, MSc. / Electrical and Computer Engineering / Symposium Talk

  • Project Summary: Morse code has historical significance in the realm of communication. While its use is not as widespread today, it still holds niche applications. One such application is in radio contesting, where radio enthusiasts compete to contact as many operators as possible within a given time window. Contacts made using Morse code are awarded more points, making it a valuable skill in this space. Although there are platforms available to learn Morse code, many of them focus primarily on individual characters. While this is a good starting point, it is essential to be able to recognize and translate longer form content such as words and sentences. This platform addresses that gap by having a language model (LLM) generate context-specific content that users must then translate from Morse code to English. This approach has a key advantage. It enables users to practice a broader range of vocabulary compared to random character databases. Moreover, the platform allows users to tailor the content to be domain-specific. For example, a user could request sentences related to naval communication or aviation. Additionally, this platform is hosted on the cloud, making it accessible to a wider audience. Users can engage with the platform anytime and anywhere, which encourages consistent practice and learning. Gamification adds an engaging aspect to the platform. Users earn more points for accurate translations. The system is designed to be scalable, with features like leaderboards, user progress tracking, and dynamic difficulty potentially being added. These features will help users gradually improve their skills. The platform could also be extended into an undergraduate-led project, provided there is sufficient interest.

   

• Seung-Yeon Jung, Ph.D. Candidate / Economics / Symposium Talk

  • Project Summary: This project investigates how government transfer programs, specifically the Pandemic Electronic Benefit Transfer (PEBT), influence grocery prices. By leveraging transaction data from a Midwest-based retailer and NielsenIQ panels, I aim to quantify the inflationary effects of PEBT-induced demand surges. Utilizing cloud computing services like AWS EC2 and S3, the project will scale demand estimation across a broad range of product categories, offering policymakers new insights into the unintended market consequences of transfer payments.

   

• Tosin Salau, Ph.D. / Department of Political Science/ Symposium Talk

  • Project Summary: My project employs a semi-supervised machine learning to extract patterns from raw textual data in news articles. To do so, I am making use of AWS S3 to store raw news reports, processed event data and then writing AWS Lambda scripts to invoke functions to test the machine learning model. Ultimately, my project seeks to develop a model that would be able to distinguish between reprisals (selective violence) and indiscriminate violence (non-selective harm to civilian populations) which is a non-trivial task.

  
   

Julie Marie Celini

Biography: Julie Celini is a second-year Master’s student in the Department of Biosystems and Agricultural Engineering at Michigan State University where she also completed her Bachelor’s degree. Julie’s interests lie in sustainable land management, particularly in the realm of agriculture, and she aspires to carry this passion into her future career. Additionally, Julie’s Spartan spirit runs deep as she proudly wore green and white while being a part of the All-Girl Cheer and STUNT Team during her undergraduate years at MSU.

Research Description: Julie’s research focuses on the improvement of agricultural practices, aiming to enhance the sustainability of agriculture while ensuring the health and fertility of the soil necessary for meeting the increasing demand for food and environmental preservation. Julie aspires to integrate this research with cloud computing, utilizing it for the analysis of bacterial and fungal communities within agricultural soil.
 

Rajarshi Pal Chowdhury

Symposium Talk

Biography: Rajarshi joined the radiation transport group in the "Experimental Systems Division" of the Facility of Rare Isotope Beams (FRIB) as a post-doctoral research associate in 2023. He has a Ph.D. in Nuclear Engineering and three years of industry experience in the areas of radiation transport and shielding for humans and electronics on the Earth and in space. At FRIB, Rajarshi focuses on contributing to the solution of complex occupational radiation shielding problems to ensure the safety of radiation workers and the general public. The career goal of Rajarshi is to be an active contributor to the radiation protection and shielding community as a researcher/educator to further human advances in the 22nd century with the safer use of radiation sources. Apart from work, Rajarshi likes to hike, travel, and is a big fan of stand-up comedies.

Research Description: In his present work, Rajarshi aims to implement applied machine learning algorithms to reduce computational costs of radiation shielding problems pertaining to high-energy accelerator facilities. His goal is to train algorithms to understand the complexity associated with radiation shielding physics to predict radiation dose more intelligibly on a relatively shorter time scale for radiation safety purposes. These models will be verified with existing established radiation transport models and/or experimental datasets that are already available.
 

Chenyang Deng

Symposium Talk

Biography: Chenyang Dang is a Ph.D. student in civil and environmental engineering at MSU. His interests include data mining and model simulation in photovoltaic systems. Chenyang is very interested in environmental sustainability and wants to be an environmental scientist in the clean energy industry. He likes music and tennis, and likes to waste a lot of time watching videos about math, medicine, history, and any subject outside his area.

Research Description: Chenyang’s research area is the modeling of end-of-life stage in solar energy. His proposed project is to collect data from utility-scale solar plants, identify the plants that have replaced the modules for energy generation, and estimate the potential waste produced by those plants.
 

Graham David Diedrich

Symposium Talk

Biography: Graham Diedrich is a graduate student at Michigan State University, pursuing a Master of Science (M.S.) in Data Science with a focus on environmental policy and climate change. His research centers on sustainable resource governance, environmental justice, and clean energy systems. Graham holds a B.A. in International Relations and a Master of Public Policy from MSU. With a commitment to advancing environmental policy through community-centered approaches, Graham strives to contribute meaningfully to creating a more sustainable and equitable future.

Research Description: For this fellowship, Graham’s plan is to focus on modeling an environmental or social system. In particular, he would like to forecast the emission reduction potential of recently proposed climate-related bills in Michigan. This scenario could be compared against a baseline business-as-usual (BAU), among others.
 

Mike Frazier

Symposium Talk

Biography: Mike Frazier is a Ph.D. student in Educational Psychology and Educational Technology (EPET) at Michigan State University and a member of The Research Laboratory for Digital Learning. He is an educator, teacher leader, and researcher with extensive experience teaching English, writing, and educational technology in public and private K-12 and higher education institutions in both the United States and South Korea. He earned his Master of Education (dual teaching certification in English, Educational Technology) from The Ohio State University and Bachelor of Arts in English from Miami University of Ohio.

Research Description: Mike studies the intersections between human and machine cognition—specifically, generative Artificial Intelligence (AI) tools like ChatGPT and their potential as learning and instructional tools. This specific project seeks to better understand the cognitive processes related to human and AI perceptions and collaborations.
 

Ishika Ghosh

Symposium Talk

Biography: Ishika Ghosh is a second-year Ph.D. student at the Department of Computational Mathematics, Science, and Engineering. She received her BSMS Dual Degree in Mathematics from the Indian Institute of Science Education and Research Tirupati (IISER-T), India. Ishika is interested in exploring the information provided by the "shape" of the datasets. Outside of academia, she interacts with the world through her camera lens and literature. She wishes to combine cloud computing and computational topology to make TDA techniques more accessib7le to fellow researchers.

Research Description: In a broader sense, Ishika’s research interest lies in applied and computational topology, especially Topological Data Analysis (TDA), which explores the hidden patterns and relationships within datasets, contributing to the advancement of data analysis techniques. She is currently working on reconstructing original data from its graph-based topological signatures.
 

Juliana Reed Burroughs Hanle

Biography: Juliana is a Ph.D. student in the Basso Lab studying remote sensing of working landscapes for natural climate solutions and atmospheric greenhouse gas reductions. She is broadly interested in computational approaches to natural climate solutions, but specifically in tracking management and identifying impacts of climate-smart land-use co-benefits. She studied history at Yale College, was a Fulbright research grantee to Norway, and has a Master’s degree in Forestry from the Yale School of the Environment. She has been working to scale sampling approaches at The Soil Inventory Project.

Research Description: Juliana is currently using Synthetic Aperture Radar, which is microwave remote sensing, to detect crop management. She is interested in leveraging cloud computing resources to remotely process and analyze satellite imagery at scale. Specifically, is it possible to swiftly and efficiently process 5-meter resolution images over seven years across the Northern Great Plains?
 

Nan Jia

Symposium Talk

Biography: Nan Jia is currently a Ph.D. student in Fish and Wildlife and Environmental Science and Policy. She spent her undergraduate and master's degrees working with RS and GIS, and her interdisciplinary background gives her a unique perspective and flexibility to use these tools for large-scale, cross-disciplinary research. Currently, she is exploring the use of big data to quantify the impacts of multiple crises and advise policymakers.

Research Description: Nan's current research area focuses on quantifying and mapping the impacts of perturbations (e.g., policies, disasters, crises, etc.) within and across systems using multiple data sources. Nan plans to use a cloud computing platform to visualize the impacts of COVID-19 on global food resilience and to reveal existing inequalities.
 

Xin Lan

Biography: Xin Lan is currently pursuing a Ph.D. in the Department of Geography, Environment, and Spatial Sciences at Michigan State University. Prior to this, he completed his master's in Earth and Environmental Engineering at Columbia University, and dual bachelor's degrees in Marine Science and Business Administration from China University of Geosciences (Beijing). Xin's research focuses on leveraging machine learning, remote sensing, and hydrological models to explore water resources. Beyond academics, he has a passion for badminton and enjoys reading in his free time. He cherishes moments spent with his partner and their cat, Woody.

Research Description: Changes in freshwater bodies can significantly impact water resources, water quality, aquatic ecosystems, and environmental health. However, there is still a lack of understanding of the large-scale spatial patterns in freshwater responses to climate change, particularly changes in freshwater coverage and temperatures for small water bodies, due to the lack of direct observations. Xin Lan's research aims to quantify changes more accurately in freshwater coverage and temperature throughout the US using Landsat data and unmixing techniques, especially for small lakes and narrow rivers, thus overcoming limitations in existing data sources.

Jiancheng Liu

Biography: Jiancheng Liu, also known as JC, is currently a first-year graduate student working under the guidance of Prof. Sijia Liu. Jiancheng’s research focuses on making AI systems both safe and scalable—something they’re truly passionate about. When they are not buried in code or academic papers, you'll find them hitting the high notes in a karaoke room or scaling a rock-climbing wall.

Research Description: Machine Unlearning is like teaching a computer to selectively forget certain things it has learned, without losing all the other knowledge it has acquired—imagine erasing just one specific memory from your brain while keeping everything else intact. Jiancheng’s project, called "UnlearnBench," is essentially a virtual testing ground that measures how well and how quickly different computer systems can perform this selective forgetting. This will help them create safer and more efficient technology that's well-suited for cloud computing platforms.
 

Fatemeh (Fatima) Fathi Niazi

Symposium Talk

Biography: Fatima is a third-year Ph.D. student in the Department of Computational Mathematics, Science, and Engineering, working under the supervision of Dr. Alex Dickson. In general, Fatima is passionate about AI, specifically generative AI models, coding, algorithm development, and engaging in multidisciplinary research. Fatima enjoys playing sports such as volleyball, soccer, and table tennis, as well as playing board games.

Research Description: In the lab, Fatima uses advanced computational techniques to simulate the interaction between drug molecules and their receptors in our bodies. Fatima’s current research focuses on applying advanced Generative AI models to a new algorithm they have developed for molecular simulation.
 

Monique Nidra Noel

Biography: Monique was born to a Caribbean lineage in Jamaica Queens, New York. She has always been an academic and leader among her peers. At the age of 16, she started community college and began to plant the seeds of a curious scientist; in 2014, she received her Bachelor of Science from the illustrious Florida Agricultural & Mechanical University. Her background is in traditional synthetic chemistry and now computational materials science is her focus. She finds it fascinating to employ computational tools to help solve future problems of the world. She enjoys sunbathing, singing, dancing, playing with her cats, watching sci-fi films, and most importantly, her family. She has always spoken for the underdog, and she looks forward to using science as a mechanism to impact the lives of the forgotten. She thanks the MSU Cloud Computing program for another opportunity to showcase her talents and bring awareness to my community! 

Research Description: Thermoelectricity is the conversion of a heat gradient into useful electricity, and vice versa. The efficiency of this phenomenon is of critical importance to the future of power generation. Zintl phases are a class of polar intermetallic compounds that combine ionic and covalent bonding to form charge-neutral structures, leading to semiconducting properties. In Monique’s research, she uses first-principles calculations to study various surface parameters of AMg2Sb2 (A=Mg, Ca, Sr) Zintl compounds, such as crystal orientation, surface energy, adsorption energy, surface reconstruction, and work function, to understand the design of these novel materials.
 

Ojo Olorunsogo Paul

Symposium Talk

Biography: Ojo Paul completed his bachelor's degree in Agriculture, specializing in Soil Science, from Ahmadu Bello University in Nigeria and began his career as a Quality Assurance Analyst at DABOL. He furthered his education with a master's degree in Soil Science from Punjab Agricultural University in India and another in Biology from Miami University in Ohio. Currently a Ph.D. student at Michigan State University, he is working on a remote sensing-based index for post-harvest corn residues under the supervision of Prof. Bruno Basso. Passionate about biking and traveling, Paul aims to become an industry leader in agro-ecosystem modeling.

Research Description: Paul’s doctoral research focuses on using remotely sensed imagery to create a new, scalable index for measuring and mapping post-harvest corn residues. By examining the spatial and temporal variability of these residues with remote sensing tools, he aims to better understand their effects on soil health and environmental feedback. Additionally, he is working on identifying proxies for Soil Organic Carbon (SOC) to incorporate them into crop models, thereby improving yield predictions through AI and process-based models.
 

Avirup Roy

Biography: Belonging to the land of the mighty Himalayas, "cloud" has fascinated Avirup a lot over the years. Engineering is his passion, music is his life. From capturing landscapes with various shapes and patterns of clouds on his camera to having delicious tea and pakoras (Indian fritters) during cloudy evenings, clouds have become an integral part of Avirup’s life. It's time for him to be closely associated with another form of cloud, now in his research.

Research Description: Learning the Megh Raga ("Raga" is a pattern of notes in Indian classical music having characteristic intervals, rhythms, and embellishments, "Megh" means cloud in Sanskrit) at a very tender age helped Avirup establish a deep connection with the clouds. Now it's his turn to convert this connection with cloud to his research which involves self-learning mechanisms on embedded hardware with computational constraints. Avirup would like to utilize cloud computing resources to overcome the computational limitations of embedded hardware devices in order to attain better self-learning capabilities.
 

Brandon Michael Webster

Symposium Talk

Biography: Brandon completed his B.S. in environmental biology at CalPoly Humboldt and has worked with plants ever since then. He is a fourth-year Ph.D. candidate in Plant Bio- Plant Breeding, Genetics, and Biotechnology (PBGB) and he likes plant breeding because it's a confluence of many different disciplines from biology to engineering. He would like to use plant breeding to help fortify and prepare agriculture for obstacles that will arise from climate change. When not at work he likes to stay active, especially bike-commuting on the river trail or playing with his two dogs. Alternatively, you can find him playing games.

Research Description: Brandon’s research goal is to better understand the biology of how maize plants respond to fertilizer input. To answer his questions, he is studying large populations of maize in the field and using remote sensing plus genomic mapping methods. These methods allow researchers to take advantage of large data sets but also come with significant computational needs.
 

Yaxuan Wang

Symposium Talk

Biography: Yaxuan Wang is a linguist with a focus on exploring the intricate nuances of language meanings, which she hopes will provide insights into key properties of the human language system. Yaxuan’s passion extends to Mongolian studies, encompassing languages, literature, history, and cultures. Beyond her professional pursuits, she finds solace in origami, painting, and exploring the rich tapestry of cultures around the world.

Research Description: Yaxuan studies how words and sentences convey meaning and how people understand them. She uses logic and experiments to figure out how language works. Additionally, she researches something fun: emojis! She looks at how emojis relate to the words people use with them.
 

Han Xu

Biography: Han Xu is a final-year Ph.D. student of Computer Science, in the Department of Computer Science and Engineering at Michigan State University. Before joining MSU, he received a master’s degree in Statistics from the University of Michigan. Han always dreams of having an academic career that he can devote to research to enhance the trustworthiness of AI techniques.

Research Description: Han has broad research interests in Trustworthy AI, including machine learning robustness, fairness, and privacy issues. He is particularly interested in studying the bias issues in adversarial robust models, and how to leverage poisoning attack techniques for data user's privacy protection. Han is also very interested in the related problems in real-world applications, such as graph data, text data, and financial data.

 ICER and DMA are proud to introduce this impressive group of fellows. Read on to learn more about each fellow and their research.

S. Farhad Abdollahi

Department/Unit: Civil Engineering

Research Description: Farhad Abdollahi's research mainly focuses on modeling the pavement structures nationwide by which he can predict the remaining service life of the pavements, different distresses on pavements, etc. These models help to evaluate the cost efficiency of different scenarios of pavement preservation, as well as estimate the associated damage cost caused by each vehicle class to the existing pavements.

Biography: Before joining Michigan State University as a Ph.D. student in the Civil and Environmental Engineering Department, Farhad gained his B.S. in Civil Engineering and M.S. in Pavement Engineering, both at Sharif University of Technology, Tehran, Iran. Farhad's main research area includes pavement analysis and design, viscoelastic materials, and computational modeling. He is currently working on a Federal Highway Administration-funded project for the modeling of numerous pavement structures nationwide.

Suhwoo Ahn

Department/Unit: Communication

Symposium Talk 

Research Description: Suhwoo Ahn's research focuses on citizens' engagement with political information and its impact on their attitudes and behaviors. Recently, he studied how bipartisan messages that emphasize common grounds between Republicans and Democrats influence the favorability of their own political party and animosity toward opposing parties. His work has been published in the Patient Education and Counseling and Journal of Political Marketing.

Biography: Suhwoo is a doctoral student in Communication at Michigan State University. He received his Master of Arts in Communication from Seoul National University and his dual Bachelor of Arts in Mass Communication and Sociology from Sogang University in the Republic of Korea. He is involved in various research projects and teaching undergraduate classes at Michigan State University.

Project Summary: This study investigates how media outlets take their partisan positions in making political news. I collected 598,998 news articles from 32 media outlets in South Korea. I will analyze them using a supervised machine learning technique based on a cloud computing service. I posit that latecomer news outlets might show stronger partisanship than established ones. I will conclude by highlighting implications for theoretical grounds and methodological advances.

Lydia Bradford

Department/Unit: Measurement and Quantitative Methods

Symposium Talk 

Research Description: Currently in Education Research, there has been a push for teacher observations during the testing of interventions to gain insights into the implementation of the intervention on the ground. Often, these observations include lengthy field notes along with a set of scoring guidelines to give the teachers scores within specified categories. Writing observation field notes and then scoring them can be very time-consuming for observers and/or researchers. Lydia Bradford's current project looks to use a variety of machine learning methods to create observation scores from the field notes themselves and then build a tool for researchers to use for scoring observations in their own research.

Biography: Lydia is a fourth-year Ph.D. student in measurement and quantitative methods with interests in statistical modeling and analysis, research design, and causal inference. She graduated from Duke University with a bachelor’s degree in Romance languages and Global Health with a minor in Economics in 2017, where she became interested in research methodology in both public health and economics. After graduating, she was a high school chemistry and economics teacher where she became interested in research methodology in Education Research which eventually led to her beginning her studies at Michigan State University. She hopes to continue research in methodology while working on larger-scale studies as a quantitative methodologist.

Project Summary: Building upon previous research (Bradford, 2022) that trained and compared different machine learning methods to score teacher observations from text to a score of 1-4, this cloud computing project creates the building blocks for an on-demand scoring system for observers using the same observation protocol. The trained machine learning model along with the necessary data preprocessing were registered through azure machine learning. A container was established and is shareable for anyone needing to access the machine learning model to score their observations. The container and access to the machine learning model has been successfully tested and is usable for the researchers currently using the PBL observation protocol.

Ritam Guha

Department/Unit: Computer Science and Engineering

Symposium Talk 

Research Description: Ritam Guha's current research focuses on predictive quality in manufacturing using time-series analysis. Every day, a huge amount of energy and effort gets wasted in manufacturing because the final product does not satisfy the quality requirement in many cases. His research work attempts to predict the quality of the final products based on the simulation of the sensor readings representing the manufacturing process and optimizes the energy utilization to reach satisfactory product quality.

Biography: Ritam received his B.E. degree in Computer Science and Engineering from Jadavpur University, India, and joined the Ph.D. program at Michigan State University after that. His research interests include evolutionary computation, deep learning, multi-objective optimization, AutoML, and many more related fields. When he is not in his lab, he can be found watching animes, playing badminton/table tennis, or driving on speedy highways.

Project Summary: Due to the ongoing global chip shortage, semiconductor manufacturing has gained massive importance in the last few years. The entire semiconductor industry is trying to scale its manufacturing capacity to cater to the needs of the global population. But currently, it suffers from huge wastage in terms of energy and resources because of the lack of recipe optimization. Semiconductor manufacturing processes are typically very long and span around 3-4 days. At the end of the process, if the product does not meet certain quality requirements, the yields are wasted and the process needs to be restarted which leads to huge wastage. So, Virtual Metrology (VM) has gained tremendous popularity as a supporting tool to optimize energy and resource utilization, thereby improving the efficiency of the manufacturing pipeline. VM refers to the automated estimation of the manufacturing properties using the data sensed from the process without physical metrology operations. We have developed a VM pipeline that takes the sensor data from one-fifth of the process run and virtually simulates the rest of the process and predicts the quality of the yields at the end. So, if the predictions do not lead to the expected properties for the yields, the process can be stopped early and restarted to save energy. As a part of the cloud computing fellowship project, I have designed and trying to deploy a web application linked to our pipeline. The application should accept the first few hours of data from the manufacturing run and estimate the final quality of the yields.

Faith Houck

Department/Unit: Kinesiology

Research Description: Faith Houck's research focuses on improving human performance in professional race car drivers. She aims to combine cloud computing and physiology data to increase human safety measures in motorsport.

Biography: Faith is a second-year Ph.D. student in the Department of Kinesiology. She works in the Spartan Motorsport Performance Lab directed by Dr. David Ferguson. Her research interests focus on improving human performance in professional race car drivers. She aims to combine cloud computing and physiology data to increase human safety measures in motorsport. When not in the lab or at a racetrack, Faith enjoys gaming, reading, or being outside with her dog.

Sue Lim

Department/Unit: Communication

Symposium Talk 

Research Description: Sue Lim's specific research area is at the intersection between human-AI communication, interpersonal communication and health communication. Sue's project will implement cloud computing to evaluate the persuasiveness of health messages generated by natural language processing (NLP) models.

Biography: Sue is a Ph.D. student in the Department of Communication at MSU. She received her B.S. degree from the Wharton School of Business at UPenn and worked as a data analyst at a marketing research company (NAXION) for four years before deciding to pursue her passion for academic research. Her general research interest is in using and developing machine learning, computational, and statistical methods to examine communication phenomena.

Project Summary: With the deployment of ChatGPT in 2022, a significant number of communication researchers have become interested in how to leverage AI for communication research. Azure cognitive services provide opportunities for communication researchers without much programming experience to use AI in their research. In this presentation, I summarize my experience with three different types of cognitive services: computer vision (spatial analysis), Azure’s bot service, and OpenAI services. Due to privacy and other constraints with the MSU Azure account, I could not continue the first two attempts . However, the OpenAI service showed promise. The second half of the presentation illustrates potential use cases of OpenAI services for communication researchers.

Sarah Manski

Department/Unit: Statistics and Probability

Symposium Talk 

Research Description: Sarah Manski's research is in Bayesian statistics. She is currently working on a project focusing on agriculture in the Midwestern U.S. The project aims to quantify the reduction in risk, especially in adverse weather conditions, associated with using regenerative agricultural practices such as rotational diversity, conservation tillage and cover cropping. She plans to integrate cloud computing into this project to expand the scope of this model to additional U.S. states, weather conditions, and soil health practices.

Biography: Sarah is beginning her fifth year in the Ph.D. program in the department of statistics and probability. She has a bachelor's degree in mathematics and computer science from Kalamazoo College and a master's degree in mathematics from Dartmouth College. Her interests are in probability theory, applied Bayesian statistics and statistics education, which are all encompassed in her current research and will form a base for her future career. Outside of academics, she enjoys studying karate, visiting her advisor's sheep farm, and spending time with her partner and their three dogs, a turtle, and a frog. 

Project Summary: In developing the largest-scale agricultural modeling effort of its kind encompassing field-level data for nearly one million fields over almost two decades with over 300 associated variables, our data storage structure has outgrown feasibility. For our researchers, using the associated project input and output data is time-consuming, memory intensive, and nearly infeasible for local machines. This project aims to create a comprehensive relational cloud database for all project data to create a reliable, scalable, and efficient storage and query environment.

Miles Roberts

Department/Unit: Genetics and Genome Sciences

Symposium Talk 

Research Description: Many studies in genetics seek to identify genetic reasons why individuals within the same species differ from each other - for example, finding genetic alterations that increase one person's risk of disease relative to another person. Miles Roberts' research project, in contrast, will be about identifying genetic differences between species that explain differences between those species. This will be useful for understanding the genetic basis of traits in species where it may be difficult to sample many individuals.

Biography: Miles is a third-year Ph.D. student in the Genetics and Genome Sciences Program. His background is primarily in evolutionary biology and he hopes to eventually apply this knowledge to a career in biotech. He spends his spare time exploring local areas and trying to discover the absolute worst movies ever made.

Project Summary: Biologists are fundamentally interested in studying the diversity of life forms on Earth, which can tell us a lot about how different forms evolved and even how they may change into the future. However, current methods for measuring diversity at the genetic level require synthesizing the inputs and outputs of many pieces of sophisticated software. For my project, I created a workflow to turn raw genetic data into diversity measurements with a single command and ran this workflow in the cloud. In the end, I produced a short tutorial on how to run workflows through Microsoft Azure’s Kubernetes Service to help other biologists make their workflows as accessible as possible.

John Salako

Department/Unit: Earth and Environment Sciences (Geological Science Major)

Symposium Talk

Research Description: John Salakos's research is aimed at understanding and predicting the spatial distribution of tree roots using geophysical tools. This non-invasive approach can reconstruct the roots of orchards (cherries, vineyards, apples, etc.). Knowledge of the spatial distribution of tree roots coupled with remote sensing imagery of the tree canopy will enhance understanding of root and shoot feedback and improve the management of orchards to better cope with climate change. He also aims to quantify carbon stocks in soils and trees in fruit orchards, linking geophysical tools with remote sensing, simulation models, and AI.

Biography: John received his bachelor's degree in Marine Science and Technology from the School of Earth and Mineral Sciences at The Federal University of Technology, Akure, Nigeria. He has working experience in a multinational company (Total Energies), where he served as a Geoscience and Reservoir and HSE Intern during his undergraduate studies. In 2020, he decided to switch carrier paths, moving from geoscience petrology to scientific computing of the earth systems ,focusing on soils and plants. He is currently a master's student in the Basso Computational Agricultural Sciences Lab under the supervision of Prof. Bruno Basso (Earth and Environmental Sciences). John aims to become a geoscience computational system modeler.

Project Summary: In this project, I aimed to optimize the runtime of a machine learning web application I created by utilizing Azure Web Application services. The primary objective was to minimize latency and improve overall performance for end users. By comparing the application’s performance on production web application services against the Dev/Test web service plan, I observed a substantial reduction in runtime delay by over 300% when using the production services. Additionally, it was found that the Azure production web services outperformed the free Streamlit cloud alternative, thus providing a more efficient solution for hosting machine learning web applications. In conclusion, leveraging Azure Web Application services for production environments or upgrading to a more robust architecture in Streamlit can significantly enhance the runtime performance of machine learning web applications.

Meicheng Shen

Department/Unit: Geography

Research Description: Meicheng Shen's current project aims to quantify the carbon and/or energy fluxes between the forest and the atmosphere. To simulate the land surface fluxes, her group drives process-based models (or statistical models) with canopy structure or leaf properties derived from multi-platform remotely sensed datasets. By studying the linkage between the forest structure and photosynthesis at the canopy scale in various climatic/abiotic conditions, she wants to understand how well they can model canopy photosynthesis with recently released remotely sensed datasets and how forest canopies may respond to the changing climate.

Biography: Meicheng is a Ph.D. student at the ERSAM lab affiliated with the Department of Geography. Her interest lies in the nexus of forest ecology and remote sensing. With more and more datasets published, she wants to explore the framework to integrate multi-source data with process-based and/or statistical models to understand how forests function as a carbon sink in the changing climate. During her spare time, she enjoys wandering in the woods or exploring historical architecture in nearby cities.

Project Summary: More and more observation datasets from the space are available to learn the land surface patterns and related ecological processes. However, processing large datasets are technical, which makes it valuable to develop sharable tools to overcome the technical barrier and advance the scientific understanding. Here, I want to develop a container that can derive canopy structural traits from discrete lidar point clouds.

Adam Terwilliger

Department/Unit: Computer Science

Research Description: Adam Terwilliger is focused on developing an AI mechanic which can understand and characterize vehicles through sound via deep learning. He also works in generative modeling and acoustic synthesis to improve data augmentation, video games, and the metaverse. He hopes his work can one day be deployed as a mobile app that anyone can use to diagnose problems with their vehicle.

Biography: Adam received bachelor's degrees in Computer Science, Mathematics, and Statistics from Grand Valley State University in 2017 and a master's degree in Computer Science from Michigan State University in 2019. He is currently a Ph.D. candidate in the DeepTech laboratory headed by Dr. Joshua Siegel. Adam will be pursuing teaching-focused positions in academia after his Ph.D. In his free time, he enjoys supporting Detroit/MSU teams, fantasy sports, board games, MOOCs, running, and biking.

Alfred Torsu

Department/Unit: Political Science

Symposium Talk 

Research Description: With a robust identification system (Ghana card) now in place, Alfred Kwadzo Torsu's research argues that the time is up for Ghana to leverage cloud computing for internet voting (I-Voting). This will not only promote faster collation and delivery of election results but also provide another medium for those who do not want to visit the polling station for a host of reasons and even for citizens who are not presently in the country.

Biography: Alfred is a second-year MPP student at Michigan State University. He is also a research assistant at Afrobarometer’s analysis unit based at MSU. His policy interest revolves around public finance, taxation, and electoral management. He holds a B.A. in Political Science from the University of Ghana, Legon. Alfred is from Ghana and intends to continue his work toward building equitable communities there and throughout Africa after he graduates.

Project Summary: The analysis of unstructured data has long been a time-consuming task for social scientists, requiring extensive reading and analysis to identify core themes and messages. However, recent advancements in cloud computing and the Azure platform have provided pre-built resources for machine learning, enabling the collection of insights from unstructured datasets with greater ease and efficiency. This project highlights the benefits of utilizing these resources, particularly in the analysis of news headlines, presidential speeches, and public policy-related tweets. By utilizing machine learning techniques, social scientists can gain valuable insights from large datasets, allowing for more informed decision-making and policy implementation.

Ziyi Xi

Department/Unit: CMSE

Symposium Talk 

Research Description: Ziyi Xi's main research area is computational seismology. He works on seismic phase (seismic signal along specific propagation path) detection using deep learning techniques in Tonga and Full-waveform tomography of East Asia and West Pacific, trying to develop refined Earth's velocity structure model.

Biography: Ziyi graduated from the University of Science and Technology of China and earned his B.S. in geophysics and B.E. in computer science. He is currently a fifth-year Ph.D. student in the department of Computational Mathematics, Science and Engineering with Dr. Shawn Wei as his mentor. He is passionate about the broad application of machine learning in science and industry and is willing to pursue related R&D roles in the future.

Project Summary: The demand for accurate and comprehensive earthquake catalogs necessitates efficient and scalable methodologies. In response, we developed a cloud computing workflow using Kubeflow on Azure Kubernetes Service (AKS) to generate earthquake catalogs from archived time-series seismograms. The workflow combines deep learning-based models with traditional geophysical algorithms, providing a scalable and cost-effective solution for processing large volumes of data. Kubeflow on Azure ensures efficient task management through Kubernetes and offers containerbased components for easy maintenance. Object storage is utilized for handling archived seismograms and temporary files, while the output earthquake catalog contains information on occurrence time, location, and phase arrival time. Our cloud-based approach demonstrates superior performance compared to traditional supercomputers in terms of efficiency and scalability. By harnessing cloud computing, this project revolutionizes earthquake catalog generation, delivering accessible and scalable solutions for seismic event analysis and contributing to our understanding of Earth’s dynamics.

Arash Yunesi

Department/Unit: Statistics and Probability

Symposium Talk 

Research Description: Arash Yunesi is working on the interdisciplinary research area between statistics, computational mathematics, and biology. Arash's group is working on single-cell RNA sequencing to identify genes and make exciting discoveries about cellular behavior using Deep Learning techniques. These discoveries will help in the near future to find more effective cures for cancer and other genetic diseases.

Biography: Before starting his Ph.D. in Statistics and Probability at MSU, Arash received another Ph.D. in Theoretical High Energy Physics from Florida State University. Even though physics will always have a special place in his heart, he has found his new research very fulfilling as it can increase the health and longevity of almost all the taxpayers that are funding his research. In his free time, Arash usually goes jogging and biking around campus and also on the Lansing River trail.

Project Summary: In this project I will demonstrate the power of cloud computing in cancer data analysis, specifically in Single Cell RNA Sequencing. The scalability and possible HIPPA compliance of cloud architecture when hosted in a secure facility enables the safe processing of individual patient data. In near future these results will enable pharmaceutical companies to produce patient-specific drugs for cancer.

• Ann Alex, Ph.D. / Institute for Quantitative Health Science & Engineering / Symposium Talk 

• Project Summary: The ENIGMA working group Organization of Imaging Genomics in Infancy (ORIGINs), a working group of investigators from around the world, is actively involved in neuroimaging research in children, particularly in infancy and early childhood. ORIGINs comprises twenty sites across six countries that collect demographic, genomic, imaging, cognitive and behavioral data. This database project is the first step towards understanding the complexities of storing ORIGINs data in a cloud-based system with the requirement of access management and the ability to compile the data for further analysis. Here, we describe the use of Microsoft Azure cloud computing resources like storage containers with encryption enabled for restricted access and Azure Databricks by means of virtual machines or compilation and analysis.

• Giovan Cholico, Ph.D. / Biochemistry & Molecular Biology / Symposium Talk

  • Project Summary: Toxicology makes use of large biological datasets, such as chromatin immunoprecipitation sequencing, bulk RNAsequencing, single-cell RNA-sequencing, and metabolomics, to investigate the role of drugs and environmental contaminants in mediating progression of diseases. These datasets are deposited into public repositories such as the National Institutes of Health’s Gene Expression Omnibus to promote data sharing and reuse. However, for researchers to obtain and reuse those datasets there are computational hurdles associated with downloading, reprocessing, and reanalyzing the datasets. The goal of this project was to implement a proof-of-concept cloud-based data commons which co-localizes the data, metadata, software, tools, and compute power to streamline data integration and reuse, and increase the potential impact of in-house generated data. Gen3 is an open-source data commons platform, which employs cloud computing services for improving data management, analysis, and shareability. As part of the Gen3 framework, Kubernetes and Terraform are used for provisioning and automated scaling of cloud resources. Moreover, microservices are used for importing data, managing user access and permissions, as well as present the Gen3 data commons through a user-friendly web-based interface. Here I describe the successes and challenges associated with deploying the Gen3 architecture using Terraform on Amazon Web Services. This effort presents a significant step forward towards establishing a multiple principal investigator use data commons to increase transparency and accessibility of research data and promote collaboration to tackle new biological questions.

• Ehsan Ashoori / Electrical and Computer Engineering

• Veronica Frans / Fisheries and Wildlife and Ecology, Evolution, and Behavior / Symposium Talk 

  • Project Summary: With increasing connections in this globalized world, human impacts on Earth’s ecosystems have become virtually unavoidable, with the most negative effects having led to our current global biodiversity crisis. In response, international governments have created multiple initiatives and targets to slow or reverse impacts on our ecosystems and promote sustainability. Among them, the United Nations ambitiously named seventeen Sustainable Development Goals (SDGs), which include Life Below Water (SDG14) and Life on Land (SDG15), and large databases have been created to monitor sustainable development progress. Such data, however, are at national scales, but most biodiversity assessment data are local or regional. It is also possible that some SDGs are being achieved at the indirect cost of the others. In effort to reduce misguidance on SDG progress for biodiversity, I am using cloud computing to create what I call the BioSDGTime database. It is a relational database that links global time series sustainable development data at national scales with biodiversity monitoring time series data at local scales. Here, I discuss the importance of cloud computing for discovering global relationships between biodiversity and sustainability.
    
     

• Xinyu Fu, Ph.D. / Plant Research Laboratory / Symposium Talk 

  • Metabolic flux analysis (MFA) is a powerful approach for quantifying intracellular metabolic reaction rates in
    plant metabolic networks. Estimation of flux parameters is solved by nonlinear fitting of a mathematical model
    describing the metabolic network to high-dimensional isotope labeling data. The rate-limiting step of MFA is the
    calculation of confidence intervals for flux parameters. Analyzing the flux confidence intervals for a leaf metabolic model with about 100 reactions takes thousands of CPU hours to complete 1000 Monte Carlo simulations. Our current workflow is enabled by submitting parallel jobs in a High-Performance Computing (HPC) environment that requires maintenance and variable queue wait times. In this talk, I will share my experiences with migrating the Monte Carlo simulation workflow from HPC to a public cloud (Microsoft Azure). Virtual Machine Scale Sets provide flexible compute capacity for running parallel simulations on an identical pool of virtual machines based on a custom image with required software (e.g., MATLAB and Python). The scalability, reproducibility, and cost of the cloud-based workflow will be discussed.
     

• Kyle Godbey, Ph.D. / Facility for Rare Isotope Beams / Symposium Talk 

  • Despite impressive technical progress in recent years, the way data is explored and interacted with in much of
    nuclear physics has remained the same. To this end we have developed the Bayesian Mass Explorer web application as a widely accessible, feature-rich tool to both display results for nuclear quantities and perform on-the-fly calculations in the cloud. The core goal for this initial phase was to produce an easy-to-use tool that researchers in both nuclear theory and experiment can navigate and integrate into their workflows, as well as providing a solid technical foundation for future development. The result is a scalable, flexible platform that allows for robust data visualization, dynamically retraining gaussian process emulators, and performing inference using pretrained machine learning models without requiring massive computational resources to be continuously allocated.
     

• Nathan Hall, Ph.D. / Department of Plant, Soil and Microbial Sciences
   

• Kirby Hermansen / Physics & Astronomy Department / Symposium Talk 

  • Project Summary: The SEparator for CApture Reactions (SECAR) is a recoil separator designed to be used in conjunction with the Facility for Rare Isotope Beams to measure nuclear reaction rates of astrophysical importance. SECAR employs a series of electromagnetic elements to filter a beam of many different radioactive nuclei into a pure beam of only the specific nuclei to be analyzed. In order to fully understand the operation of such a complex system as SECAR, experimentalists use COSY Infinity simulations to model the behavior of a beam through SECAR. However, with a functionally infinite magnetic field phase space for each of the nineteen electromagnetic elements, scanning the possible parameters requires the use of supercomputing clusters or cloud-based computing networks. Our cloud-based pipeline connects COSY simulations with a multi-objective evolutionary algorithm, MOEA/D, in order to identify optimal SECAR settings given specific input parameters. The pipeline is self-contained as an Azure Container Instance for simplicity and ease of deployment. The simulated results from all container instances are collected and
    stored in a central location for potential future uses.
     

• Nan Jia / Fisheries and Wildlife
   

• Herve Kashongwe / Department of Geography / Symposium Talk 

  • Project Summary: Tropical forests are key global ecosystems for sequestering carbon, stored primarily as above ground biomass (AGB, units: t/ha) and are thought to account for about half of global terrestrial biomass carbon and more than a third of the current terrestrial carbon sink. They support high levels of biodiversity and regulate global climate with teleconnections to other parts of the world. Inventories of tropical forests and AGB are imprecise and sparse. Increasingly, Light Detection and Ranging (LiDAR) and optical wavelength satellite data are used to map and monitor forests and estimate forest AGB and carbon stocks. My PhD research seeks to evaluate whether the new generation of satellite data provided by the NASA Harmonized Landsat-8 Sentinel-2 (HLS) and the Global Ecosystem Dynamics Investigation spaceborne LiDAR, utilized in conjunction with airborne and ground data, can improve
    tropical forest AGB and carbon estimation. In this experiment, cloud computing is used to scale my research to larger domains in space, specifically from mapping forest canopy height and AGB of a 11,000 km2
    study site in Mai Ndombe province, DRC to mapping the entire Mai Ndombe province (∼127,000 km2) forest canopy height and AGB.

• Rabin Kc / Plant, Soil, and Microbial Sciences
   

• Shruti Khandelwal / Urban and Regional Planning / Symposium Talk

  • Project Summary: Food security has been directly associated with the ability to access foods; however, this data is scarce and does not include data about availability of culturally appropriate foods for ethnic minority groups. This project aimed to involve a visual mapping simulation of data for accessing an online hosted food security map. Simultaneously, the map represents the score of food access of culturally appropriate foods updated from online survey in real time at specific geographic locations. The web-based platform is easy for ethnic minority groups to engage with, even for those with limited internet resources. 

• Wei Liu, Ph.D. / Department of Geography / Symposium Talk 

  • Project Summary: Studies have shown that exposure to green spaces is associated with better mental health. Exposure to green spaces does not only include physical accesses, but also the visibility of green space within sight of a neighborhood. However, green spaces such as parks are not evenly distributed in an urban setting and the view of a green space may be blocked by buildings. Thus some neighborhoods could have limited visibility of green spaces even when a park is within a relatively short distance. This study aims to use cloud computing to perform green space visibility analysis to calculate green space visibility scores at a human-eye level from the road networks in Detroit. Outcomes from this analysis can be used to test the association between visibility scores and health outcomes

• Xiaorui Liu / Department of Computer Science and Engineering / Symposium Talk 

  • Project Summary: Distributed machine learning (ML) aims to develop distributed optimization algorithms to speedup computationally expensive problems in machine learning. However, it is unclear how the system configurations impact the efficiency and scalability of the algorithms. In this project, we try to take the advantages of cloud computing to make a comprehensive efficiency benchmark. Specifically, I will measure the time cost and cost decomposition of popular distributed algorithms under varying factors such as the number of computing devices and model size. We hope this can provide deep insights about the efficiency bottleneck in distributed ML.

• Krishna Pothugunta / Information Technology and Management / Symposium Talk 

• Project Summary: Post-2013, Sharing Economy (SE) platforms have grown tremendously and are estimated to grow from $14 billion in 2014 to $335 billion by 2025. Airbnb and Uber are two major players in the SE sector. Why did the valuation of Airbnb increase so much even when the bookings were canceled and dropped? This study is intended to extend prior work by including other amenities such as transportation hubs, restaurants, stores and tourist attractions and their impact on the survival of the Airbnb listings. The study uses Python (Haversine metric) to identify the distance between the listing and other amenities for a particular location. However, Python takes a lot of time to calculate the distances on a local machine, and so Data Analytics Spark is employed on the Azure cloud server to increase the processing speed. The intended outcomes proved to be more effective when the SQL scripts are used instead of Python on Spark.

• Guangjing Wang / Computer Science / Symposium Talk 

  • Project Summary: With the proliferation of smart devices and the growing capability for automation, smart homes are becoming increasingly popular. In a smart home, it is common that devices and applications from different platforms interact with each other. As a result, there will be complex system behaviors that are hard to diagnose, especially for heterogeneous closed-source smart home platforms. Existing work proposes to collect the provenance of device data using code instrumentation. However, most smart home platforms are closed-source, which limits code instrumentation methods. The existing provenance methods are hard to put into practice for heterogeneous closed-source platforms. Nevertheless, event logs and application descriptions are both available and essential sources for checking devices’
    functionalities and states, which can be fused for building provenance graphs. However, the event logs in a home contain private information, which should not be shared with a third party to track threats or train models. In this project, we use federated graph neural networks to implement threat detection in smart home provenance graphs. Meanwhile, we can keep users’ data locally to protect users’ daily data privacy. We apply virtual machines in Azure to simulate compute clients in a federated framework. The main research contribution is to solve the federated graph neural network over a non-independent and identically distributed dataset and then a threat detection model for interaction threats in smart homes.

• Kaidi Wu / Department of Economics / Symposium Talk 

  • Project Summary: The opioid epidemic has played an important role in contributing to the death toll of drug overdose deaths in recent years. According to the Centers for Disease Control and Prevention, about 840,000 people have died since 1999 from a drug overdose, of which 70% involved an opioid. The deaths were not only driven by illicit drugs, but also from legitimate sources like prescribed opioids. To ease the rapid increase in overdose death involving opioids, states have imposed various regulations on opioid prescription. Two of the most popular policies are the prescription drug monitoring program (PDMP) and the opioid prescription limit laws. The PDMP requires doctors to look up prescription history before they can prescribe opioids. The prescription limit laws limit the quantity of opioids that can be prescribed in the patient’s first visit. In this paper, I investigate the causal impact of these policies on opioid-related deaths and its secondary effect on crime with individual level mortality data and incident level crime data respectively. Using quasi-experimental approaches, I provide evidence that policies with the same good intent can lead to different results due to various measures taken. The results suggest reducing access to prescribed opioids when people are already addicted drives people to resort to illicit market and is inefficient in addressing overdose deaths in the short run; while the prescription limit laws helped to reduce deaths and do not have a negative impact on crime. This study also contributes to the existing literature by showing that the evaluation of policies can be biased if it does not account for the interactions of different policies.

Symposium Talk

• Dr. Mohamed Abouhawwash / Institute for Quantitative Health Science & Engineering

• Donald Akanga / Geography & Spatial Sciences 

• Julie Butler / Physics & Astronomy

• Yingying Chen / Advertising and Public Relations

• Rachel Domagalski / Mathematics

• Harrison Fernandez / Computational Mathematics, Science, and Engineering

• Beth Gerstner / Fisheries & Wildlife

• Mimi Gong / Fisheries & Wildlife

• Dr. Jared Homola / Fisheries & Wildlife 

• Jing Kong / College of Business

• Caleb Lucas / Political Science

• Dr. Chris Mancuso / Biochemistry & Molecular Biology 

• Viviana Oritzlondono / Field Crop Pathology 

• Dr. Mahmoud Parvizi / Computational Genomics

• Dr. Ruijuan Tan / Plant Breeding, Genetics, and Biotechnology

• Dr. Rachel Toczydlowski / Integrative Biology

• Dr. Tong Zhou / Computational Mathematics, Science, and Engineering