Restructure CNAI Landscape to AI Native Landscape

[ChaoyiHuang]

Restructure the landscape to better support modern AI Native workloads, organizing projects into four workload categories with a unified AI Native Infra category as foundation layer:

• AI Agent (e.g., Agent Framework, State and Memory)
• Inference (e.g., Framework, Runtime)
• Training (e.g., Pre-Training, Post-Training)
• Data (e.g., Data Architecture, Data Science)
• AI Native Infra (e.g., Gateway, Orchestration and Scheduling, Workload Runtime, Accelerator and Superpod)

The new structure reflects the evolving tech stack, particularly the rise of AI Agents and multimodal foundation models, while building on CNAI's strong foundation. The infrastructure layer abstracts underlying resources to support diverse workload requirements.

The landscape adopts a three-layer structure to better visualize both technical hierarchy and workflow relationships, as refined through community discussions:

Category Organization with Vertical Hierarchy and Horizontal Pipeline (Technical Stack):

1. Base Layer: AI Native Infra (foundational infrastructure)
2. Middle Layer: Horizontal Workflow Pipeline (Data → Training → Inference)
3. Top Layer: AI Agent(end-user facing applications)

Subcategory Organization:

• Within each category, subcategories are ordered top-to-bottom
• Sequential numbering indicates logical dependencies where applicable

This structure emerged from multiple rounds of community feedback balancing technical accuracy with visual efficiency.

Credits:

• Includes refinements from in-person discussions with Vincent Caldeira and Omri Shiv
• Inspired in part by the Agentic Community landscape structure [1]
• Incorporates insights from CNCF community discussions:
  • AI TCG channel [2][3]
  • Landscapers channel [4][5]

[1] https://github.com/agentic-community/agentic-landscape
[2] https://cloud-native.slack.com/archives/C08Q78J65A7/p1775118381260299
[3] https://cloud-native.slack.com/archives/C08Q78J65A7/p1772614288421669
[4] https://cloud-native.slack.com/archives/CM09QERF1/p1773198799247499
[5] https://cloud-native.slack.com/archives/CM09QERF1/p1774949382211369

Pre-submission checklist:

Please check each of these after submitting your pull request:

• Are you only including a repo_url if your project is 100% open source? If so, you need to pick the single best GitHub repository for your project, not a GitHub organization.
• Is your project closed source or, if it is open source, does your project have at least 300 GitHub stars?
• Have you picked the single best (existing) category for your project?
• Does it follow the other guidelines from the new entries section?
• Have you added your SVG to hosted_logos and referenced it there?
• Does your logo clearly state the name of the project/product and follow the other guidelines?
• Does your project/product name match the text on the logo?
• Have you verified that the Crunchbase data for your organization is correct (including headquarters and LinkedIn)?

[github-actions]

You can preview your changes by visiting this link.

Note

This feature is still experimental and may not work as expected in some cases. Please report any issues you find!

[cynthia-sg]

For reference, you can also see how these changes look when integrated with the new settings proposed in the landscape2-sites repo here.

[ChaoyiHuang]

@caniszczyk @jeefy @mrbobbytables @tegioz @cynthia-sg

Following community discussion on the ai native landscape layout, achieving the desired horizontal category layout (Data/Training/Inference in a row, and in the middle layer) requires a new feature from landscape2 (Ref: landscape2#941).

The reason for the such a layout can be found in the PR's description.

To move forward, the implementation strategy could follow one of these three paths

• Merge as-is: Proceed with the current layout (which slightly differs from the community conclusion) and ignore the row container feature.

• Two-step update: Merge the current landscape.yml and settings.yml first. Once the landscape2 feature is released, we can update settings.yml to reflect the discussed layout.

• Hold for feature: Keep these two PRs open and wait for the landscape2 feature. We will then configure settings.yml to the final version and merge everything together.

[omrishiv]

If the changes to landscape are not accepted, I would suggest reordering as such:

agentic
inference
training
data
infra

[caniszczyk]

I would love to move forward with this initially and deal with the row layout later (it doesn't hinder from the experience imho)

[caniszczyk]

@ChaoyiHuang amazing work!

[ChaoyiHuang]

I would love to move forward with this initially and deal with the row layout later (it doesn't hinder from the experience imho)

Thank you all. @caniszczyk @jeefy @mrbobbytables @tegioz @cynthia-sg This approch works.