[GSoC 2026] Inquiry regarding Project 1: Agentic RAG on Kubeflow

[Rakshanda Thakur]

Dear Kubeflow Maintainers,

I am Rakshanda Thakur, an MS Computer Science student at California State University, San Bernardino (graduating May 2026), specializing in AI/ML systems and data engineering. I am writing to express my strong interest in pursuing Project 1: Agentic RAG on Kubeflow as a Large (350-hour) GSoC 2026 project.

I have been studying the kubeflow/docs-agent repository and the project's vision of evolving the current RAG implementation into a robust agentic architecture. My background aligns closely with the technical challenges of this project:

• Production RAG Experience: I built and deployed an AI survey chatbot at CSUSB using LangChain, FAISS, and Sentence Transformers, implementing semantic + thematic retrieval over institutional Oracle databases with real-time subgroup filtering.
• Agentic & LLM Experience: I fine-tuned and integrated Llama models with LangChain orchestration for a university RecWell chatbot, including RAG pipelines, prompt optimization, and hallucination reduction.
• Deployment & Infrastructure: I have built production-grade Docker multi-stage deployments with security scanning (Docker Scout, Burp Suite) and CI/CD pipelines following Agile methodologies.
• Backend Engineering: I architected a FastAPI/Uvicorn backend integrated with MySQL and Oracle for a web-based scheduling system, including role-based workflows and real-time dashboard features.
I have already introduced myself on the CNCF Slack (#kubeflow-contributors) and have begun exploring the codebase. I am currently working on two entry-point contributions:
• Issue #92 — Adding Milvus resource constraints and WSL2 memory prerequisites to the local Kind setup documentation
• Issue #81 — Deduplicating repeated text-cleaning regex patterns into a shared utility module to improve code maintainability
I hold an Oracle Cloud Infrastructure 2025 Generative AI Professional certification, which aligns directly with the OCI deployment component of this project.
I would greatly appreciate any guidance on additional technical areas or entry-point issues the mentors would recommend to help me demonstrate proficiency with the docs-agent architecture before submitting my proposal.

Thank you for your time and for the opportunity to contribute to the Kubeflow ecosystem.

Best regards,
Rakshanda Thakur
MS Computer Science | CSUSB
https://github.com/Rakshanda3
https://www.linkedin.com/in/rakshanda-thakur-29a502207/