Hackathon project 2026 In progress

Inpatient Chart Brief

An AI context layer that turns years of patient history into a sourced, glanceable handoff brief a clinician can absorb in under a minute.

Product, interaction, and coded MVP

AI / Clinical IntelligenceHealthcare0→1 Solo Build

A resident gets paged about a patient they’ve never met. It’s 2am, and there is about 60 seconds before walking into the room. The chart holds the answer but buries it across years of notes. Inpatient Chart Brief reads the whole chart and surfaces the few things that matter right now—every claim linked to the source note it came from. No source, no bullet.

Inpatient Chart Brief presenting prioritized clinical context beside source documentation — The brief: a concise clinical synthesis beside the source documentation that supports it.

See the build

The full walkthrough—from the chart-review haystack to a sourced clinical brief.

What makes it different

Relevance over completenessThree to five prioritized signals, not a ranked dump of the chart. The product judgment is as much about what to leave out as what to include.
Provenance, not confidenceNo decorative confidence meters. Every insight expands to the real, dated note it came from, making attribution the trust mechanism.
Honest about limits“Nothing to flag” is a first-class state. Stale evidence is labeled rather than hidden, and the product is explicit about what it does not know.

I’m designing and building Inpatient Chart Brief end to end—product, interaction, and the coded MVP—with a practicing pediatric-neurology resident as the source-of-truth user. It is a bet on where AI belongs in high-stakes work: a calm, sourced synthesis delivered at the moment of need.