Feedback: Improving Developer Experience with Data Science

Qovery leverages a data-driven approach to enhance the reliability and simplicity of application deployments, accommodating diverse languages and frameworks while addressing common issues through automation and machine learning (ML). By collecting error data during deployments, Qovery employs an ETL pipeline to extract, transform, and load data, preparing it for analysis using Python and natural language processing (NLP) tools. The data is processed to remove unnecessary elements and then vectorized using Doc2Vec for clustering with DBSCAN, which helps identify prevalent error patterns. Insights from this process reveal that a significant portion of issues arise from build process errors, Kubernetes misconfigurations, and Docker image build failures. This structured analysis allows Qovery to prioritize and address these challenges efficiently, ultimately improving deployment stability and enabling continuous refinement of their platform's performance.