Speaker
Description
Scientific software often grows out of necessity. Researchers build scripts, notebooks, and custom tools to solve immediate problems, but these workflows can become difficult to maintain, reproduce, and support. In this talk, I will discuss how our small software team works with scientists to introduce software engineering practices without disrupting how they work.
Rather than asking scientists to completely change their workflows, we build around them. We introduced Git and self hosted Gitea through lightweight versioning approaches, including Git submodules, allowing researchers to continue developing their code while we handle integration into larger applications. We have also used AI tools to help translate between scientific code and production software.
Once integrated, we treat scientific software as an external component that we can support without fully owning. We use containers for challenging environments such as IDL and Fortran on Macs, CI/CD pipelines for deployment, and services such as Celery to connect scientific tools with web applications. This allows us to add error tracking, reporting, and monitoring while helping scientists improve the reliability and accessibility of their software.
I will share lessons learned from turning informal scientific workflows into sustainable systems while preserving the flexibility that made them useful in the first place.