Grid AI ensures continuous improvement for in-house code quality with DeepSource
Founded by the creators of PyTorch Lightning, Grid AI is a platform for training models that enables rapid research iteration. It aims to simplify scalable AI research so that when a network becomes complex, code doesn’t.
Python, Terraform, Go
"DeepSource, we never turn it off. We feel like it's finally a quality tool that we can turn on and never turn off. And it just keeps fixing things for you. And it just keeps getting better over time too."
— Timothy Chen, Software Engineer
Grid AI was rapidly growing, which meant that the team had to constantly revise their code. This drew a high risk of bugs getting introduced to their codebase. The team was struggling to set up a code quality and security DevOps infrastructure from scratch. Using open-source tools such as Pylint wasn't helping the case because they don't run all the time and don't help with catching the most important issues. They tried other tools to help catch issues but they were very noisy.
The solution required a tool that wasn't noisy — this is where DeepSource helped. The new engineering team was able to easily adopt DeepSource and streamline their code review workflow with zero friction. DeepSource integrated seamlessly with their Python and Terraform repositories on GitHub, and the Analyzers were up and running within minutes.
4x growth in analysis runs in 4 quarters
Grid AI started analyzing repositories crucial to their product in the July of 2020, with just 77 runs in the first month. The Analyzer impressed the team with the vast categories of issues it was able to detect, with a false positive rate of lower than 5%.
Fast forward to 2021, the engineers of Grid AI have consistently increased their reliance on DeepSource Analyzers.
Just Autofix it!
DeepSource's Autofix feature suggests fixes for issues detected and creates pull requests with the recommended changes. The team at Grid AI was overworked with fixing commonly occurring issues, and so it adopted this feature to their workflow in the last quarter of 2020. "Generating code fixes is a game-changer for us, to not having to spend all this time. I wish Autofix just exists in my IDE", said Timothy. Grid AI almost doubled its usage of Autofix by end of the first quarter of 2021.
An improved code review workflow
DeepSource has constantly acted like a foremost quality gate for Grid AI's pull requests. The team always has DeepSource Analyzers running in the background. As DeepSource continues to generate fixes for the most important issues occurring in Grid's repositories, it has helped expedite and improve their code review workflow on a daily basis.
Shift left, enterprise-grade.
Start building with the most sophisticated static analysis platform for your workflow and prevent bugs before they end up in production.
Deploy on-premise to have absolute control of your data
Onboard thousands of repositories in minutes, not months
Save over 4 hours on average per developer every week