FIND-20260323-016 · 2026-03-23 · Innovation Veille
MediaPipe Solutions — On-Device AI Inference for Web, Android, iOS and Python
adhoc
MEDIUM
MediaPipe Solutions (Google) is a cross-platform suite of on-device AI tasks covering vision (object detection, face/hand/pose landmarks, image segmentation, gesture recognition), text (embeddings, classification, language detection), audio classification, and generative AI (LLM inference, RAG, function calling). Models are ~16 MB, Apache-2.0 licensed, support ONNX, and run locally without any cloud dependency. SDKs available for Android, Web (WASM), iOS, and Python. A Model Maker tool allows fine-tuning on custom data. Flagged via Slack @neural_avb tweet (2026-03-23).
Source
https://ai.google.dev/edge/mediapipe/solutions/guide
ODS Impact
Directly applicable to ODS product layer. DocSign (Tauri/React/Rust desktop) and OSIGN SaaS could integrate the Web or Python SDK for: (1) KYC liveness detection using face landmark models — critical for African fintech compliance; (2) document boundary detection/segmentation before PDF upload in DocStore; (3) handwritten signature detection in the PDF Engine. The Web SDK runs inside a browser or Tauri WebView with no backend roundtrip, preserving data privacy. Python SDK enables server-side prototyping. Apache-2.0 license is fully compatible with ODS stack. No integration blocker identified.
Security Review
License: Apache-2.0 | Maintenance: ACTIVE | Risk: LOW | Recommendation: SAFE_TO_USE
Tags
ai
on-device
edge-inference
mediapipe
google
onnx
kyc
vision
web
python
tauri
docsign