All organizations · GSoC 2026 · Infrastructure and cloud, Artificial Intelligence

Kubeflow

The Machine Learning Toolkit for Kubernetes (AI)

Our goal is to streamline the process of scaling and deploying machine learning (ML) models to production by leveraging Kubernetes' strengths: <br /><br /> - Seamless deployments: Effortless, repeatable, and portable deployments across diverse infrastructure, allowing you to experiment on local machines and then seamlessly transition to on-premises clusters or the cloud. <br /><br /> - Microservice management: Efficient deployment and management of loosely-coupled microservices for building modular and scalable ML pipelines. <br /><br /> - Dynamic scaling: Automated scaling based on demand, ensuring optimal resource utilization. <br /><br /> User-centric design: Recognizing the diverse tool preferences within the ML community, we prioritize user customization (within reasonable boundaries). Our system takes care of the repetitive tasks, freeing users to focus on the core ML challenges. <br /><br /> Starting focused, expanding rapidly: While we initially focused on specific technologies, we actively collaborate with various projects to integrate additional tools and broaden our reach. <br /><br /> Our vision: A future where simple manifests empower you to utilize a user-friendly ML …

  • python
  • go
  • kubernetes
  • typescript
  • YAML
  • machine learning
  • kubernetes
  • AI/ML
  • generative AI

Project ideas

11 ideas on its 2026 ideas page; of the 11 that state a difficulty: 2 medium, 9 hard. Sizes: 2 medium, 9 large. 16 mentors named.

  1. Project 1: Agentic RAG on Kubeflow (Expansion of kubeflow/docs-agent) large hard
  2. Project 2: OptimizationJob CRD for Hyperparameter Optimization large hard
  3. Project 4: Platform Scalability and Security large hard
  4. Project 5: Helm Charts large hard
  5. Project 6: MCP Server for Kubeflow SDK medium medium
  6. Project 7 : Integrate Kubeflow SDK with OpenTelemetry large hard
  7. Project 10: Dynamic LLM Trainer Framework for Kubeflow large hard
  8. Project 11: Composable Kale Notebooks with Visual Pipeline Editor large hard
  9. Project 12: Kubeflow SDK/SparkClient - Batch Jobs, Observability & Production Readiness large hard
  10. Project 13: Kubeflow Pipelines - Enhance Support for Pod Lifecycle Failure medium medium
  11. Project #: End-to-End ARM64 Support & Validation on Kubeflow large hard

Read the ideas page. Ideas are for scope reference; never copy one into a proposal.

History

In GSoC since 2020, 4 years. Projects per year: '20: 2, '24: 8, '25: 11, '26: 9.

On GitHub

25 repositories pushed to in the last year. Main languages: Python (10), Go (7), Jupyter Notebook (2), TypeScript (2), HTML (1), YAML (1).

Links and channels

go deeper in your terminal
git-into-gsoc org "Kubeflow" --shortlist
git-into-gsoc repos "Kubeflow"
git-into-gsoc issues --org "Kubeflow"
or ask your assistant:
"Which Kubeflow repositories merge
newcomers' pull requests, and which beginner
issues are still free?"

Those commands read live data: which repositories merge outside pull requests, and which beginner issues nobody has claimed. Install in two minutes.