Tale

Your Software is Served
Carol Herre joined SBGrid in early 2014, joining the internal staff at Harvard Medical School as a release engineer. She has taken on the challenge of applying the SBGrid model to other fields, such as genomics, and is spending her time finding, compiling and testing new software.

“We’re trying to provide the biological scientific community with software to do their jobs,” she says. “The existing model works really well for structural biology, so we’re trying to extend that to other biological sciences.”

Herre is working with the Tools-and-Technology Committee at HMS to select applications, most of which are software programs that are already available on the Orchestra Cluster, a computing cluster at HMS. “These are popular, so that’s where I’m starting,” she says.

Career-wise, Herre began as a naturalist with a forestry degree from Penn State University. The work was rewarding, but isolating, so she started to take other classes. Before she knew it, she was back in school, this time at East Stroudsburg University earning a Master’s degree in computer science.

While Herre initially worked providing computer support for a small university, she later moved on to work for Thinking Machines, where she helped users of super computers solve their software problems.

Her other experience includes working on configuration management and web development as a contractor for a variety of clients. But her main qualifications are personal. “I’m a bit obsessive and really try to think about how things will look to a user,” she says.

One of the challenges Herre is facing as she prepares RNA sequencing and genome processing tools for distribution is that each piece of software she touches has a different set of underlying dependencies. “I really want to make it a push-button experience for the user,” says Herre, who worries that the alternative — a multi-step user experience requiring documentation that is easily overlooked and quickly outdated — would be a mediocre solution.

The other challenge is validation. These tools are complex and scientifically advanced, with some computational tasks taking days to complete, so validating that an installation produces correct results is not always feasible. “I have to be thinking about validating the quality of the installation, and not the quality of the tool,” she says.

Herre is working closely with the team that manages the Orchestra Cluster to learn about their experiences providing the tools to users. As time goes by, she hopes to begin working more closely with users and developers. She’s also trying to get a better handle on the science. “I am working on it, little by little,” she says.

--Elizabeth Dougherty