Harmony

STRUCTURED DATA, CITED TO THE CHART.

Harmony turns unstructured oncology records into registry-ready and evidence-grade structured data. Same OncoLLM under the hood, configured to produce auditable outputs against NAACCR, CoC, QOPI, or any custom data dictionary you bring. Every field links back to the line in the chart that produced it.

Extraction accuracy

Faster case abstraction time

Data Fields Accepted AS-IS

Faster case finding

Why this matters

The gold standard for structured oncology data is still a person reading hundreds of pages by hand.

Cancer registries, real-world evidence programs, and precision oncology projects all run on structured data abstracted from unstructured charts. The work is tedious, error-prone, expensive, and depends on a clinically trained workforce that’s burning out. Every year the volume of charts to curate grows faster than the people to curate them.

What it does

Two workflows.
One platform.

Each workflow is measured, and built around the work oncology teams do every day.

Cancer registries, on auto-pilot.

Harmony automates curation of all NAACCR data elements from unstructured data. Implements NAACCR standards, dictionaries, custom collection forms, and CoC reporting requirements out of the box. Registrars cut abstraction time by an estimated 75%. Integrations with leading registry systems of record auto-populate the cases, so the work that remains is verification.

Automated curation of every NAACCR element in unstructured data

~75% reduction in case abstraction time

Native support for NAACCR, CoC reporting, custom forms

Auto-population into leading registry systems

Evidence-grade data for Observational Research or RWD.

Results from our preprint on structuring unstructured medical records. Read the preprint →

94% accuracy across 100+ data elements vs. expert-curated benchmarks

95% acceptance rate from expert abstractors

Minimal edits across 15,000+ data points

Trust

Verify any field, in seconds.

Every Harmony output is two pieces: the structured value, and a citation that opens the source line in the chart. Registrars and abstractors stop hunting for evidence and start verifying it. CoC reviewers and sponsor auditors get an end-to-end trail without extra reporting infrastructure.

Any dictionary

Tune Harmony to your project in weeks.

Beyond NAACCR, Harmony tunes per project on your data dictionary, for precision oncology cohorts, RWE protocols, registry add-ons, or sponsor-specific extraction tasks. Typical tuning timeline: a few weeks of lead time from receipt of dictionary to production-ready agent.