An open data initiative by RxAI

Every medicine should be recognizable.

The open, always-current visual identity layer for medicines—declared by manufacturers, validated in pharmacies, and built for machines.

Manufacturer-declared Pharmacy-validated Open by design
01 / The gap

Drug data is essential infrastructure. Visual identity data is still treated like a gated add-on.

Commercial databases are broad and deeply embedded—but redistribution is restricted, source evidence is hard to inspect, and real-world image variation rarely becomes shared infrastructure. The last federal pill-identification service was retired in 2021.

THE PUBLIC IMAGE GAP / 2020 ARCHIVE 9,779 of 83,925 archived Pillbox appearance records were flagged as having an image.
11.7%

Historical benchmark, not current market coverage. NLM’s final dataset is static and explicitly not for present-day pill identification. Source ↗

01

Access is expensive

Public contracts show institutional drug-data spend in the tens of thousands per year and above.

Verified procurement
02

Updates are vendor-shaped

Daily and weekly feeds exist, but customers cannot inspect or extend the full change pipeline.

Closed process
03

One image is not a distribution

A polished reference cannot represent lot, tooling, lighting, wear, or camera variation.

ML coverage gap
02 / How it works

From declared product to validated appearance.

Every asset keeps its origin, evidence, freshness, and confidence. Synthetic references extend the dataset; they never masquerade as observed truth.

01
DECLARED

Manufacturer feed

Imprint logic, product mappings, calibrated images, effective dates, and change events.

02
OBSERVED

Pharmacy validation

Consented field images cluster into valid appearance modes; outliers go to review.

03
DERIVED

Synthetic coverage

Traceable variants cover lighting, angle, occlusion, cameras, and safe hard negatives.

{

Nothing loses its lineage. Raw capture, normalized crop, canonical mode, and synthetic derivative remain linked through immutable asset IDs.

origin · parent_ids · captured_at · model_version · license · confidence }
03 / For manufacturers

You already publish the label. Now publish the truth behind the image.

We can seed product facts from DailyMed. A direct change feed makes your products easier to identify, safer to dispense, and faster to support—without waiting for a closed redistribution cycle.

  • 01 Control effective dates and superseded appearances
  • 02 Publish imprint symbols without flattening them into text
  • 03 See field variation and correction evidence
  • 04 Grant clear rights for public use and ML evaluation
Become a founding contributor
04 / For developers

A visual identity API with evidence attached.

Query product identity, search ambiguous imprints, or request reference assets without hiding how a result became trustworthy.

IDENTITY ASSETS
Illustrative API response
GET https://api.rxid.example/v1/ndc/00000-1234-10 200 OK
{
  "ndc": "00000-1234-10",
  "rxcui": "demo-1045",
  "identity": {
    "imprint": { "obverse": "A 14", "reverse": null },
    "shape": "OVAL",
    "color": ["PALE_MINT"]
  },
  "evidence": {
    "status": "validated",
    "observed_images": 42,
    "appearance_modes": 2,
    "confidence": 0.997
  },
  "updated_at": "2026-08-24T14:32:11Z"
}
manufacturer observed synthetic Read the proposed schema ↗
05 / The first release

A useful core, then a compounding loop.

Phase 1 targets — not live counts
0+ imprint-indexed NDC products

Seeded from current public SPL records and normalized to product-level identity.

0+ real reference images

Manufacturer and licensed public-source assets with origin and capture metadata.

0+ synthetic training variants

Clearly labeled derivatives for camera, lighting, orientation, and occlusion coverage.

0% assets with provenance

No unlabeled mixing of declared, observed, or generated visual evidence.

ASSET ORIGINEvery image says what it is.
Manufacturerdeclared
Pharmacyobserved
Generatedsynthetic
06 / Roadmap

Ship the evidence layer first.

We start where openness creates immediate utility, then earn the right to expand.

  1. NOW → Q4

    Seed + specification

    SPL ingestion, open schema, licensing review, archived-source audit.

    In design
  2. Q1 2027

    Founding cohort

    Three manufacturer feeds, pharmacy capture pilot, public correction workflow.

    Partner
  3. Q2 2027

    Open beta

    Versioned bulk release, public API, search explorer, evaluation set.

    Build
  4. LATER

    Certified infrastructure

    Enterprise SLAs, verified feeds, global expansion, governance independence.

    Scale
07 / Working names

Built like a public resource. Named like one.

Creative directions only—trademark and domain screening still required.

  1. 01RxID CommonsCurrent working name
  2. 02OpenImprintLiteral + focused
  3. 03DoseprintVisual fingerprint
  4. 04Tablet CommonsWarm + open
  5. 05OpenDose RegistryInstitutional
  6. 06PillprintApproachable
RXAI / OPEN INFRASTRUCTURE FOUNDING COHORT / 2026

Help make every medicine legible—to people and machines.

We’re looking for manufacturer data stewards, pharmacy image partners, clinical safety reviewers, and developers to shape the first open release.

Read the research brief
Concept site: connect this CTA to your preferred RxAI form or email before launch.