Summer’s Up

PSC Interns Advance AI, Cybersecurity, Computer Administration

Sept. 1, 2017

On two dates in July and August 2017, PSC’s summer interns—a central part of the center’s educational mission—reported the results of their projects to the PSC staff. The projects, which ranged from physical chemistry to computer science to biology, forwarded the interns’ learning and provided advances to PSC user and staff research.

In a “deep learning” artificial intelligence (AI) project leveraging the Bridges system, Ishtar Nyawira and Kristi Bushman, students at the University of Pittsburgh, expanded a 2016 intern project involving Nyawira as well as Annie Zhang and Iris Qian, both from Carnegie Mellon University, to help researchers at Harvard University and PSC create a 3D map of the connections of the zebrafish larva brain. (See the results of the larger zebra connectome project here.) Their aim was to use AI to automate the painstaking manual task of segmenting electron microscope images of ultra-thin slices of zebrafish brains, a necessary step in re-assembling those image slices into a 3D map. The ultimate goal is to speed the process by which neuroscientists create a 3D view of the brain’s connections so they can more rapidly begin asking scientific questions about what those connections mean for brain function.

Other projects undertaken by the 2017 interns include:

  • Generating reports and data that help PSC staff manage the center’s supercomputers, better serving users’ needs.
  • Creating better models for release and control of neurotransmitters at the junction between nerve and muscle cells in the frog, including a visualization of how the signal is transferred from the nervous system to movement.
  • Improving accuracy of the computation of interactions between the nucleus and core electrons in metal atoms.
  • Testing cybersecurity tools for authenticating users in electronic systems, identifying security vulnerabilities and backing up systems so that they are resistant to “ransomware.”
  • Automating the conversion of MRI scan images into an interactive 3D model of the brain, to aid in brain tumor diagnosis and surgical planning.