[GSoC 2025 PROPOSAL] Yuxiang Jiang - Jenkins-specific LLM for CI Failure Analysis

Contributing GSoC
1

Hi Jenkins community :waving_hand:,

My name is Yuxiang(Ryan) Jiang, and I’m currently pursuing a controller’s in Computer and Information Science at Cornell University. I’m passionate about combining machine learning, developer tools, and systems optimization to solve practical engineering problems.

For GSoC 2025, I’ve submitted a proposal titled “Fine-tuning a Jenkins-Specific LLM for CI Failure Analysis.” This project is based on the official idea “Domain-specific LLM based on Jenkins usage data” and aims to build an intelligent assistant that can help Jenkins users diagnose build and test failures using a fine-tuned LLM or Retrieval-Augmented Generation.

The proposal outlines:

  • A data pipeline for processing real CI logs from ci.jenkins.io
  • Fine-tuning or RAG integration using LLaMA 2 or similar open-source models
  • A web-based assistant (React + Django/FastAPI)
  • Target use-case: classifying failures as infra issues, flaky tests, or actual bugs

I’ve reviewed the 2024 Jenkins LLM project, set up a Jenkins instance locally, and started participating in the Gitter channel. I’m excited about contributing to Jenkins and exploring how AI can improve CI/CD developer experience.

:memo: Proposal Attachment：
You can view my full proposal here on Google Docs:
:link: Fine-tuning a Jenkins-Specific LLM for CI Failure Analysis
The document is publicly viewable, and feedback or comments are very welcome!

Looking forward to learning from the community and collaborating with mentors or contributors who have worked on related projects!

Thanks!
Yuxiang

1 Like