Hey team! 👋 I'm Guna, a software engineer with experience in Python, Java, Kubernetes, and LLM pipelines (open-source contributor to Haystack and Jenkins). I'm really interested in contributing to Metaflow for GSoC 2026. I've set up the dev environment and dug into the Client API internals. I'm excited about the "Agent-Friendly Metaflow Client" project — having built RAG pipelines in Haystack, I have firsthand experience with how LLM agents struggle with inefficient metadata retrieval. Would love pointers on where to start the audit. Thanks!