@OpenMed_AI

Open-source AI for healthcare. 2,000+ Apache-2.0 clinical models. 600+ MLX and 700+ ONNX variants. Built for local and on-device AI. Free, forever.

github.com/maziyarpanahi/openmed
Joined August 2025
  • Tweets2.4K
  • Following12
  • Followers8.1K
395Media
OpenMed 3.0 is out: Patient Journey. One synthetic patient, five systems: a clinic note, hospital FHIR, an outside clinic's HL7, a lab CSV, a retina DICOM report. They disagree. 3.0 links every fact to its exact source, keeps the conflict open and amends without erasing.
7
31
1
357
15,657
The access check in code: one boundary, two views. A researcher asking for the identified view gets role_denied. The de-identified view answers without a subject id. Withdraw consent and the same clinician request comes back consent_withdrawn.
2
8
660
Docs: github.com/maziyarpanahi/ope… Runs offline, no hosted policy service. Synthetic data.

openmed/docs/compliance/separated-projections.md at master · maziyarpanahi/openmed

Local-first healthcare AI: clinical NER and HIPAA PII de-identification on hardware you control. 2,200+ medical models, 35 model-backed PII languages, and Python, MLX, Android and browser runtimes....

github.com
116
Same patient, same question, three answers. Clinician: the identified record. Researcher: refused on the identified view, served the de-identified one. Consent withdrawn: nothing, for anyone. OpenMed 3.0 audits every attempt, and the audit log holds none of the content.
2
2
19
903
Separated projections in openmed 3.0.0: github.com/maziyarpanahi/ope… Synthetic journey. The demo script wrote the de-identified view; the SDK keeps both apart, gates each read and writes a content-free audit row. Apache 2.0.

openmed/docs/compliance/separated-projections.md at master · maziyarpanahi/openmed

Local-first healthcare AI: clinical NER and HIPAA PII de-identification on hardware you control. 2,200+ medical models, 35 model-backed PII languages, and Python, MLX, Android and browser runtimes....

github.com
2
139
Three mentions, three honest outcomes: mapped, ambiguous, unmapped. "hypertensive urgency" stays unmapped even though I10 (hypertension) is in the vocabulary. Close is not the same as correct.
1
4
586
Docs: github.com/maziyarpanahi/ope… The mention text is stored as a keyed hash, not as text. Synthetic demo subset.

openmed/docs/clinical/terminology-resolution.md at master · maziyarpanahi/openmed

Local-first healthcare AI: clinical NER and HIPAA PII de-identification on hardware you control. 2,200+ medical models, 35 model-backed PII languages, and Python, MLX, Android and browser runtimes....

github.com
1
155
In a note, "PE" can mean pulmonary embolism or pleural effusion. OpenMed 3.0 returns both codes, I26.99 and J90, marks the mention ambiguous and sends it to review. It does not pick one. A term with no match stays unmapped instead of being forced onto the nearest code.
1
1
17
816
Terminology resolution in openmed 3.0.0: github.com/maziyarpanahi/ope… Synthetic 3-concept ICD-10-CM subset. OpenMed bundles no vocabularies; you bring your own. Apache 2.0.

openmed/docs/clinical/terminology-resolution.md at master · maziyarpanahi/openmed

Local-first healthcare AI: clinical NER and HIPAA PII de-identification on hardware you control. 2,200+ medical models, 35 model-backed PII languages, and Python, MLX, Android and browser runtimes....

github.com
3
162
Hugging Face's science team counts OpenMed among the most active AI-for-science contributors, with Ai2 and NVIDIA. 2,277 open models and 26 open datasets for medicine: huggingface.co/OpenMed Open weights and open data are how medical AI earns trust.

OpenMed (OpenMed)

healthcare AI, clinical NER, biomedical NLP, PII / de-identification, on-device AI, edge inference, Apple Silicon / MLX, data sovereignty, privacy-preserving ML, open-source

huggingface.co
Emma Scharfman
@EmmaScharfmannOct 16
Hugging Face is becoming more and more the home of AI x Science projects. Open source accelerates scientific discovery when models, datasets, and tools are shared openly, all researchers can reproduce them and build on top of them. A big thanks to @allen_ai, @OpenMed_AI, and @NVIDIAAI for being the most active contributors of AI for Science artifacts on @huggingface these days! 🤗 Hugging science heatmap: huggingface.co/spaces/huggin…
2
3
62
2,169
huggingface.co/OpenMed

OpenMed (OpenMed)

healthcare AI, clinical NER, biomedical NLP, PII / de-identification, on-device AI, edge inference, Apple Silicon / MLX, data sovereignty, privacy-preserving ML, open-source

huggingface.co
98
The FHIR round trip in code. Export a 2-fact slice and the loss report names what did not fit: resolution_target_unsupported. Import it back and the facts are identical. Edit one status in transit and you get fhir_custody_conflict.
1
1
14
952
Docs: github.com/maziyarpanahi/ope… Local and offline: no terminology server, no hosted validator.

openmed/docs/interop/fhir-cohort-roundtrip.md at master · maziyarpanahi/openmed

Local-first healthcare AI: clinical NER and HIPAA PII de-identification on hardware you control. 2,200+ medical models, 35 model-backed PII languages, and Python, MLX, Android and browser runtimes....

github.com
2
119
A patient journey to FHIR R4 and back: 6 facts out as 8 entries, 6 identical facts back. Now flip one Condition from inactive to active in transit. OpenMed 3.0 refuses the import as a custody conflict instead of merging it. Every trip also reports what it could not carry.
2
15
863
FHIR Journey round trips in openmed 3.0.0: github.com/maziyarpanahi/ope… Synthetic record, 5 sources. The added resources and the edit were made by our test script. Apache 2.0.

openmed/docs/interop/fhir-cohort-roundtrip.md at master · maziyarpanahi/openmed

Local-first healthcare AI: clinical NER and HIPAA PII de-identification on hardware you control. 2,200+ medical models, 35 model-backed PII languages, and Python, MLX, Android and browser runtimes....

github.com
112
Liquid's new LFM2.5-24B-A2B did a full skin follow-up review in 19 seconds, for under half a cent. 4 visits, synthetic patient, 5 tool calls, 0 failed. It caught the spot that grew 1.29 mm and cited the exact measurement for every finding. 2B active. Strong release @liquidai
5
15
1
166
7,433
Our first take wrote that L02 was "showing no growth." We threw it out. −0.05 mm inside ±0.20 mm of measurement error isn't no growth. It's "can't tell yet." Tightened the instruction, re-ran, kept both runs. Clinicians: how do you word this in a follow-up note?
1
183
The spot that grew: 8.70, 8.98, 9.40, 9.99 mm. +1.29 mm vs ±0.20 mm measurement error, so it's flagged. L02 moved −0.05 mm, inside the error, and it got that right. LFM never sees the images. OpenDerm measures, LFM pulls the history via tool calls and writes the cited review.
1
2
353
It also left L07 open. It has no verified match in earlier visits, so there's no trend to report, and it didn't invent one. Open weights on @huggingface: huggingface.co/LiquidAI/LFM2… Synthetic images, hosted run, model wait sped up 3x in the video. A review aid, not a diagnosis.

LiquidAI/LFM2.5-24B-A2B · Hugging Face

We’re on a journey to advance and democratize artificial intelligence through open source and open science.

huggingface.co
3
226
Ask OpenMed 3.0 for a clinical brief from a model that isn't on the machine, and you get nothing back. No quiet cloud fallback. The deterministic local path returned a brief where every sentence cites an exact span of the note, with contact details masked first.
2
1
32
1,137
Guarded clinical briefs: github.com/maziyarpanahi/ope… Synthetic walkthrough with fixture NER/NLI. The brief contract is synthetic-only in 3.0, not a qualified summarizer for real patients.

openmed/docs/clinical/clinical-brief.md at master · maziyarpanahi/openmed

Local-first healthcare AI: clinical NER and HIPAA PII de-identification on hardware you control. 2,200+ medical models, 35 model-backed PII languages, and Python, MLX, Android and browser runtimes....

github.com
2
170
Load more