Anne S Berres
Anne S Berres
My main research focus is on data science at scale. My work includes data workflows, data fusion, analysis, and visualization on different platforms from web applications to high-performance computers. As a postdoc in the Computational Urban Sciences group, I work primarily with large geographic and urban data. You can learn more about my current work in the projects section.
I currently serve as Treasurer for the Oak Ridge Postdoctoral Association (ORPA). Last Fiscal Year, I was the Research Committee chair for ORPA. In this role, I organized multiple research seminars, created the new Your Science in a Nutshell (a workshop on science communication), and organized the Annual ORPA Research Symposium.
Before my time at ORNL, I was a postdoc in the Data Science at Scale group at Los Alamos National Laboratory where I worked on analyzing the connection between use of computational resources and cognitive value of the outcome by compressing image databases and evaluating the usability. I furthermore worked as a developer for MPAS-Ocean, the ocean component of the Energy Exascale Earth System Model (E3SM) project. During that time I extended multiple analysis members to enable the use of regional masks.
I completed my BSc (2009), MSc (2011), and PhD (2015) at the University of Kaiserslautern in Kaiserslautern, Germany. During my studies, I specialized in visualization, image processing, and computer vision. My minors were mathematics (BSc) and biology (MSc).
Urban Exascale Computing Project (UrbanECP)
The goal of UrbanECP is to couple simulations for urban climate, population, and traffic and run them at scale. My contribution to this project is in coupling simulations, directing the data flow between them, and analyzing their outputs.
Real-Time Data and Simulation for Optimizing Regional Mobility in the United States
Real-Time Mobility Control System
There are many projects, such as the Vaccine Delivery Project by the Bill and Melinda Gates Foundation, which require the knowledge of up-to-date population counts in remote areas. However, census information may be unavailable or outdated. Instead, this information is gained from satellite imagery.
Detecting human settlements in satellite imagery is a tedious task when done manually by humans. Machine learning can help accelerate this process.
Accelerated Global Human Settlement Discovery
Accelerated Global Human Settlement Discovery is a subproject of the Settlement Mapping project, which is aimed at porting the detection process from smaller machines like local clusters or DGX-1 boxes to leadership class computing facilities. The team was granted 25 million core hours on Titan, which even after 5 years, is still one of the top 10 computers world-wide. My contribution to this project has been in the development of a scalable image workflow, parallelized image-preprocessing to feed into deep learners, and management of the project.
Energy Water Nexus Knowledge Discovery Framework (EWNKDF)
Energy and Water systems depend on each other, and this interdependence is of great interest to the energy sector, as well as energy customers. There are hundreds of different data sources, all with different data types, formats, and access. The Energy Water Nexus Knowledge Discovery Framework is a webtool to pull these different sources under the same umbrella and provide tools to make them accessible for analysis and comparisons between datasets. My contribution on this project extends from user interface design, data aggregation, processing, and fusion, to visualization, and includes the integration of WebWorldWind as an alternate map to OpenLayers into the WSTAMP system.
- Titan, SummitDev (OLCF)
- Theta (ALCF)
- VMs (CADES)
- Globus (OLCF, CADES, ALCF)