@dry-beach-38304 @square-wire-39606 I am writing to express my strong interest in the GSoC project “Sandboxed Execution Environments with Devcontainers” for Metaflow. The idea of enabling reproducible, isolated execution of workflows locally—without relying on cloud infrastructure—is highly compelling and aligns closely with my interests in scalable systems and developer tooling. 🧑‍💻 Background I am Ayush, a 2nd-year undergraduate student in Mechanical Engineering at IIITDM Jabalpur. Alongside my core studies, I have been actively working in Python, Machine Learning, and system-level development, with a growing focus on reproducible pipelines and containerized environments. 🔧 Relevant Skills & Experience • Python & System Development: Built ML and data pipeline projects • Docker: Experience containerizing applications and managing environments • AI/ML Systems: Worked on RAG-based systems and applied ML projects • Open Source: Contributed PRs involving bug fixes, testing, and feature improvements I am comfortable working with tools that involve environment isolation, reproducibility, and automation, which are central to this project. 🎯 Why This Project & My Approach This project stands out to me because it solves a real gap in Metaflow: enabling local, reproducible, and sandboxed execution without requiring Kubernetes or cloud infrastructure. From my understanding, key challenges include: • Seamless integration of the Devcontainer specification into Metaflow • Managing container lifecycle during step execution • Ensuring reproducibility across different local environments • Maintaining compatibility with existing Metaflow decorators 💡 Initial Approach 1. Design

@devcontainer

Decorator Extend Metaflow’s execution model Allow users to specify a devcontainer configuration 2. Execution Engine Use Docker to spin up devcontainers locally Execute each step inside an isolated environment 3. Environment Handling Support

.devcontainer/devcontainer.json

Ensure dependency reproducibility 4. Integration & Testing Ensure compatibility with local and CI workflows Add tests for isolation and reproducibility 🚀 What I’ve Already Done Explored Metaflow’s architecture and decorator-based execution model • Studied the Devcontainer specification and its usage in VS Code/Codespaces • Worked with Docker for local environment isolation • Built small prototypes involving containerized execution workflows • Started reviewing how Metaflow handles step execution and environments I am ready to: • Set up Metaflow locally and experiment with flows • Start exploring relevant issues and contributing incrementally • Dive deeper into implementation details