Code reviews are one of the most critical — yet time-consuming — parts of the software development process. As teams grow, maintaining consistent review quality and catching issues early becomes increasingly difficult.
In this talk, we share our experience from a real-world experiment where an AI-powered code review, costing just $0.44, identified a potential issue during the development process. This experience highlighted an opportunity to further improve how and when code reviews take place.
We will present how we integrated AI-assisted code reviews into our GitLab CI/CD pipelines using Codex, enabling automated feedback directly within merge requests. The goal was not to replace human reviewers, but to support developers by identifying potential issues earlier and improving overall code quality.
The presentation will walk through the architecture of the solution, how AI analysis is triggered in the pipeline, and how feedback is surfaced back to developers. We will also share lessons from real usage, including challenges such as false positives and developer trust.
Key takeaways:
In this talk, we share our experience from a real-world experiment where an AI-powered code review, costing just $0.44, identified a potential issue during the development process. This experience highlighted an opportunity to further improve how and when code reviews take place.
We will present how we integrated AI-assisted code reviews into our GitLab CI/CD pipelines using Codex, enabling automated feedback directly within merge requests. The goal was not to replace human reviewers, but to support developers by identifying potential issues earlier and improving overall code quality.
The presentation will walk through the architecture of the solution, how AI analysis is triggered in the pipeline, and how feedback is surfaced back to developers. We will also share lessons from real usage, including challenges such as false positives and developer trust.
Key takeaways:
- Integrating AI-assisted code reviews into CI/CD
- Practical architecture for GitLab workflows
- Lessons learned from real-world usage
Dimitris Kyriakopoulos
KOTSOVOLOS
Dimitris Kyriakopoulos leads the Node.js Chapter at Kotsovolos Digital Factory, where he and his team build and operate large-scale microservice systems powering a modern e-commerce platform.
His work focuses on backend architecture, distributed systems, and improving engineering practices across development teams, with particular emphasis on reliability, observability, and developer productivity.
Recently, his team has been exploring practical ways to integrate AI into the software development lifecycle. One of the initiatives they introduced is AI-assisted code reviews integrated into GitLab CI/CD pipelines, helping developers receive automated feedback during merge requests and identify potential issues earlier in the development process.
Dimitris enjoys sharing real-world engineering lessons from operating production systems and exploring how AI can enhance developer workflows.
His work focuses on backend architecture, distributed systems, and improving engineering practices across development teams, with particular emphasis on reliability, observability, and developer productivity.
Recently, his team has been exploring practical ways to integrate AI into the software development lifecycle. One of the initiatives they introduced is AI-assisted code reviews integrated into GitLab CI/CD pipelines, helping developers receive automated feedback during merge requests and identify potential issues earlier in the development process.
Dimitris enjoys sharing real-world engineering lessons from operating production systems and exploring how AI can enhance developer workflows.