Review and expand AI patterns with the Green AI Committee

[russelltrow]

Context

Many of the existing AI patterns were written by Navveen in 2022 and need updating. The Green AI Committee and Software Standards Working Group are natural homes for this work — they have the domain expertise that the Patterns project team does not. Additionally, new patterns are needed for emerging areas like Agentic AI.

Navveen has a separate repository of AI patterns that could be incorporated or referenced. He also flagged that AI/ML-specific lifecycle phases (training, inference, fine-tuning) don't map neatly to the SDLC-based top-level categories, which needs to be resolved.

What needs to happen

• Russ to share the current AI patterns with the Green AI Committee
• Green AI Committee to review existing AI patterns: what's still relevant, what needs updating, what's missing
• Navveen to review the new persona-based categorisation against AI/ML patterns and flag where training, inference, and other ML-specific phases should sit
• Explore incorporating Navveen's separately published AI patterns repository into the catalogue (or referencing them)
• Identify and draft new patterns for Agentic AI before the SCI for AI launch
• Ensure AI patterns include the new properties being added to the catalogue (impact indicators, author attribution, persona tags)

Acceptance criteria

• All existing AI patterns reviewed and either confirmed, updated, or retired
• At least 3 new AI patterns added (including Agentic AI coverage)
• AI/ML lifecycle phases are clearly navigable within the new category structure (via subcategories, tags, or both)
• Patterns are ready for inclusion in the catalogue ahead of SCI for AI publication

Dependencies

• Categorisation structure finalised (Update README.md #1)
• New pattern properties defined (patterns for storage volume optimisation #6)
• SCI for AI timeline (Adding list of Tags #4)

[russelltrow]

Link to CNCF White Paper - [WHITEPAPER] Sustainable AI in Cloud-Native Environments

[russelltrow]

Notes from the GAIC meeting on 1st September:

Current State

The Green Software Patterns catalogue was created 2-3 years ago as a repository for best practices while the GSF focused on developing the SCI standard. It has collected approximately 100+ patterns from member organizations but suffers from poor usability due to its technology-based organization (AI, Cloud, Web categories). A 2024 workshop concluded the current structure is not useful for practitioners trying to find relevant guidance for their specific roles.

Restructuring Initiative

The project is transitioning to a persona-based approach with 13 identified roles, from AI/ML Engineers to Product Managers and UX Designers. Each pattern will be tagged to relevant personas and include structured information: problem description, solution, SCI impact, assumptions, considerations, and references. This approach aims to help practitioners quickly identify patterns relevant to their specific responsibilities.

Key Challenges

• Project co-leads (Francesca and Liya) lack AI domain expertise
• Need subject matter experts to own and maintain persona categories
• Existing patterns require updating given rapid field evolution
• Gap between technical patterns and behavioral/organizational guidance needed

Collaboration Opportunities

Vincent from Red Hat shared CNCF's parallel work on cloud-native AI optimization patterns organized around AI lifecycle stages (data, training, inference). Rather than duplicate efforts, GSF could reference CNCF's technical work while focusing on higher-level strategic practices. This collaboration could provide detailed technical patterns while GSF addresses broader organizational and behavioral aspects.

Actions

Immediate (Next 2 weeks)

• All GAIC members: Complete the card sort survey to help categorize existing patterns
• Russ: Share survey link and meeting materials with committee members
• Vincent: Provide specific links to CNCF documentation and join upcoming patterns project calls

Short-term (1-2 months)

• GAIC members with AI expertise: Consider taking ownership of AI/ML Engineer persona category
• Francesca/Liya: Schedule follow-up calls with Vincent to explore CNCF collaboration details
• Project team: Audit existing AI patterns against current best practices and update outdated content