This website is for the DS5110 Big Data Systems course at the University of Virginia by Dr. Judy Fox. This course covers the current state and future trends in hardware and software for Big Data Systems and applications. Our overarching goal is to build efficient software and hardware for big data processing, analytics, and machine learning at scale. The site is rendered on GitHub Pages and comprises student projects for the class.
Problem Statement: Modern Astronomical Research requires Large-scale image data processing and prediction. This is important for giving astronomers further insight into the origins of the universe as well as its evolution. The large size of the models required for this is very computationally and resource-intensive.
Solution: The proposed solution for this problem is the Cloud-based Astronomy Inference (CAI) framework. This is a scalable framework leveraging serverless computing with AWS. CAI integrates pre-trained foundation models with a Function-as-a-Service (FaaS) Message Interface (FMI), enabling efficient and accessible inference on astronomical images. This approach minimizes the computational burden on individual devices by utilizing distributed processing and cloud resources. Students used this cutting-edge framework to analyze a preprocessed astronomy dataset and test results and viability.
Figure: CAI Framework Architecture
Results: Overall students found that using CAI showed good performance and was also cost-effective at the same time.
Collaborators:
Team 1: Dan Anthony, Harold Haugen, George Shoriz, Zack Lisman
Team 2: Isidro Pride, Abhinandan Mekap, John Le
Team 3: Karthika Solai, Aqsa Majeed, Kanitta Srichan, Chi Do
Team 4: Ryan Healy, Nicholas Miller, Charles Lotane