Leveraging OpenInfra and Open Source GenAI to Address Climate Change

In an era where AI is reshaping software development, the OpenInfra community stands at the forefront of integrating foundational models (FMs) and AI agents into open-source workflows. This talk delves into how OpenInfra projects can capitalize on these technologies to address pressing challenges such as contributor productivity, energy efficiency, and climate change.

By utilizing datasets from OpenInfra projects and adhering to OpenInfra workflows, I've conducted experiments within the OpenStack and Hugging Face ecosystems. I've fine-tuned a suite of open-source models—including LLaMA 3, CodeT5, SantaCoder, Mistral-7B, and OpenLLaMA-13B—to assist developers and contributors in a variety of socio-technical activities. These activities encompass micro-collaboration, documentation enhancement, code review optimization, test case generation, event planning, predictive analytics in CI/CD pipelines, and automation—all within the context of OpenInfra projects.

The result is a significant boost in contributor productivity across more than 21 distinct use cases, demonstrating the tangible benefits of integrating GenAI into open-source development. This talk will provide the OpenInfra community with practical insights and best practices for leveraging foundation models to accelerate development, improve sustainability, and address climate change effectively.