E.P.I.C. Pipeline Optimizer

Inspiration

As an AI/ML student aiming to build sustainable tech, I realized that CI/CD pipelines run thousands of times daily, wasting massive energy on bloated Docker images and redundant jobs. I wanted to build a digital teammate that tackles this exact bottleneck and makes green computing the default.

What it does

It acts as a sustainable digital teammate within GitLab. When I want to check a pipeline, the agentic chat automatically reads my .gitlab-ci.yml file, spots inefficiencies (like bloated node:latest images or missing caches), and suggests optimized code (like lightweight Alpine images) directly in the chat, aiming to reduce execution time by 40-60%.

How I built it

I built it using the GitLab Duo Agent Platform. I created custom YAML configuration files to define the agent's identity and gave it access to tools like read_file. I wrote a highly specific system prompt tailored to identify energy-wasting CI/CD patterns and used the Agentic chat interface to trigger the file analysis.

Challenges I ran into

Initially, my agent configuration failed the GitLab validation pipeline because I included custom triggers not supported by the strict sandbox schema. I had to learn how to adapt my YAML file to strictly adhere to the allowed public: true and system_prompt schema and rely on the agentic chat for the trigger instead.

Accomplishments that I'm proud of

I'm proud of successfully deploying a functioning GitLab Duo Agent that provides mathematically sound optimizations (like an 80% size reduction via Alpine images) and getting the green checkmark on my validation pipeline!

What I learned

I learned the ins and outs of the GitLab Duo Agent Platform, how to format strict YAML configurations for AI agents, and how continuous integration pipelines are structured (and how they waste compute power if not carefully managed).

What's next for E.P.I.C. Pipeline Optimizer

I plan to expand the agent's capabilities to automatically benchmark the exact carbon footprint of a pipeline run and suggest the most energy-efficient server regions for deployments.

Built With

• gitlab-ci/cd
• gitlab-duo
• yaml