Skip to content

feat: add Edge prediction pattern#48

Open
Technocrat-dev wants to merge 1 commit intomercari:masterfrom
Technocrat-dev:feat/edge-prediction-pattern
Open

feat: add Edge prediction pattern#48
Technocrat-dev wants to merge 1 commit intomercari:masterfrom
Technocrat-dev:feat/edge-prediction-pattern

Conversation

@Technocrat-dev
Copy link
Copy Markdown

What this PR does / why we need it:

Adds the Edge prediction pattern, which was listed as "To do" in the README. This pattern covers running ML inference directly on edge devices like phones, IoT hardware and embedded systems.

Partially addresses #1 (covers the inference pattern, future PRs can add edge training and model management patterns)

Changes:

  • New Serving-patterns/Edge-prediction-pattern/design_en.md with full pattern doc
  • New architecture diagram
  • Updated README.md to link to the new pattern instead of "To do"

What the pattern covers:

  • Usecases for on-device inference (low connectivity, latency requirements, privacy)
  • Architecture: model optimization (quantization, pruning), OTA distribution, edge deployment tiers (micro/thin/thick)
  • Pros: low latency, offline support, privacy, bandwidth savings
  • Cons: limited compute, update complexity, device fragmentation
  • Considerations: monitoring, security, cloud fallback

Special notes for your reviewer:

I followed the same format used by the other Serving patterns. Happy to adjust the content or add more detail if needed. This is my first content contribution to this repo.

Add the Edge prediction pattern to the Serving patterns category.
This pattern describes running ML inference directly on edge devices
such as smartphones, IoT devices and embedded systems.
Covers model optimization (quantization, pruning), edge-cloud model
distribution via OTA updates, and considerations for monitoring,
security and fallback strategies.
Closes mercari#1
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant