skills.rest
maziyarpanahi
maziyarpanahiCOMMUNITY
Explore all skills by this author →
Website𝕏@maziyarpanahiGitHub

running-openmed-ondevice

Deploy OpenMed clinical NLP models locally on Apple Silicon, iOS, or browser backends.

5.0k|615|Updated Oct 4, 2025
Search Tags:#mlx#webgpu#onnx#coreml#on-device ai#clinical nlp#openmed
One-click install
npx skills add https://github.com/maziyarpanahi/openmed --skill running-openmed-ondevice
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: running-openmed-ondevice
Source: https://github.com/maziyarpanahi/openmed/tree/main/skills/running-openmed-ondevice
Command: npx skills add https://github.com/maziyarpanahi/openmed --skill running-openmed-ondevice

SYSTEM DOCUMENTATION & REQUIREMENTS

What problem does it solve?

This skill helps deploy OpenMed clinical NLP and language models locally, avoiding network exposure and keeping sensitive patient data on hardware you control.

Core Features & Use Cases

  • Backend Selection: Choose MLX for Apple Silicon, CoreML for iOS/macOS Neural Engine deployments, or ONNX/WebGPU for cross-platform and browser inference.
  • Model Conversion and Quantization: Convert Hugging Face token-classification models into MLX, CoreML, or ONNX artifacts and optimize them with 4-bit, 8-bit, int8, or fp16 quantization.
  • Local Clinical Inference: Run NER, de-identification, and MLX language generation entirely offline, including local artifact loading through the standard OpenMed API.
  • Validation and Safety: Benchmark quantized models against full-precision versions, verify ONNX exports, account for platform limitations, and protect cached PHI and model outputs.
  • Use Case: Deploy a de-identification model in an iPad application or run privacy-preserving clinical text analysis on an Apple Silicon workstation without uploading patient data.

Quick Start

Use this skill to choose an on-device backend, convert and optionally quantize an OpenMed model, then run and validate it locally for clinical NER or de-identification.

Frequently Asked Questions about running-openmed-ondevice

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run clinical NLP models locally without sending patient data to a network service?▼

Running clinical NLP models locally requires deploying OpenMed models entirely on your hardware using backends like Apple Silicon MLX, CoreML, or ONNX/WebGPU to process NER and de-identification tasks offline without network exposure.

Can I convert Hugging Face token-classification models for on-device inference on Apple Silicon?▼

Yes, converting Hugging Face token-classification models into MLX artifacts allows you to run on-device inference on Apple Silicon, optimizing clinical text analysis workloads with local artifact loading through the standard OpenMed API.

Does CoreML support offline de-identification for iOS and macOS applications?▼

CoreML supports offline de-identification for iOS and macOS applications by leveraging the Neural Engine to run OpenMed clinical NLP models locally, ensuring sensitive patient data never leaves the device hardware you control.

What is the best way to quantize OpenMed models for cross-platform browser inference?▼

The best way to quantize OpenMed models for browser inference is exporting to ONNX and applying 4-bit, 8-bit, int8, or fp16 quantization, enabling cross-platform WebGPU workflows for privacy-preserving clinical text analysis.

How do I validate quantized clinical models against full-precision versions for accuracy?▼

Validating quantized clinical models against full-precision versions requires benchmarking locally, verifying ONNX exports, accounting for platform limitations, and protecting cached PHI and model outputs to ensure clinical safety standards.

Why should I not use cloud APIs for clinical text analysis and de-identification?▼

Using cloud APIs for clinical text analysis risks exposing sensitive patient data to network services, whereas deploying OpenMed models locally on your own hardware avoids network exposure and maintains strict data privacy compliance.

Related Skills

Explore More Software Engineering Skills→
mgd34msumgd34msu
4

huggingface-js

Integrate Hugging Face models into JavaScript apps via Transformers.js and the Inference API.

Community
Intermediate
maziyarpanahimaziyarpanahi
5.0k

authoring-model-cards

Convert OpenMed clinical evaluation reports into governance-ready model cards.

Community
Advanced
maziyarpanahimaziyarpanahi
5.0k

building-with-openmed

Guide developers in building local-first clinical NLP pipelines with OpenMed.

Community
Advanced
maziyarpanahimaziyarpanahi
5.0k

deploying-openmed-mcp

Expose clinical NLP tools via MCP server over stdio or HTTP.

Community
Advanced