Navya plans expert cancer treatment for hospitals. We built the pipeline that gets a patient from raw hospital records to a structured, evidence-backed case summary, and the interface the clinicians actually work in. It runs in production.
to a clinician-ready case summary, down from days
EHR intake through summary delivery, one pipeline
the interface designed with the people who work in it
THE PROBLEM
Every case started with a pile nobody could read fast enough
A new case arrived as records from a hospital system, and no two hospitals sent the same thing in the same shape. Genetics, pathology, imaging, prior treatment and how the patient responded to it were all in there somewhere, in documents built for storage rather than for reading.
Someone had to read all of it before a case could even be considered. That put days between a patient and a plan, and it capped the work at the number of people available to do the reading.
THE PIPELINE
From hospital EHR to a structured case, automatically
We built the integration layer that pulls records in from each hospital system and normalises them into one shape, so the rest of the pipeline stops caring where a case came from.
Behind it, processing and classification models label what is actually in the documents: the disease picture, the genetics, the treatments tried and the response to each. Research and summary automations then do the reading that used to be manual, and the whole thing is deployed as one continuous flow rather than a set of steps somebody runs by hand.
THE SUMMARY
The genetic picture, generated for patient intake
Intake is where a case is won or lost, and genetics is the part that takes longest to assemble by hand. Generating that summary was a long build: getting it right matters more than getting it fast, because a clinician has to be able to trust it.
The output is a structured summary a clinician can act on, with the evidence behind each line traceable back to the source document rather than asserted.
THE INTERFACE
Designed with the clinicians who use it
A summary nobody trusts is a summary nobody uses, so the interface was designed with the clinicians working in it rather than handed to them. Every generated line keeps its supporting evidence one click away.
The result is an end-to-end product, from hospital integration through to the screen a clinician reads, running in production.
Also delivered at Navya
One pipeline, from hospital integration through to the clinician’s screen.
- EHR integration across hospital systems, normalising records into a single shape
- Data processing, labelling and classification pipelines
- Genetic summary generation for patient intake
- Research automations for evidence gathering
- Summary automations replacing manual record reading
- A clinician-facing interface designed with its users
- End-to-end deployment, from integration through to production
Discuss your next system.
Start with a free, one-hour engineering review. Leave with a practical next step.