The clinical intelligence layer

The method

LLM-assisted workflows search, organise, and extract published literature. The collection brings population studies, case reports, and mechanistic biology into a common research framework. It currently contains 13,476 sources. We label evidence maturity and record findings that qualify, reframe, or contradict a concept. Automated extraction can be wrong; links to the original studies are essential for checking the work.

What the research can and cannot establish

Case reports are valuable for noticing unusual presentations and generating hypotheses. They do not establish how common a finding is, prove causation, or validate a treatment. Study counts describe the collection, not certainty. A verified biological mechanism does not mean the full Zinda framework has been externally or clinically validated.

Our research programme also includes model evaluations, the Gene Atlas, and structured evidence extraction. Internal benchmarks are exploratory. Private protocols, training records, and evaluation cases remain outside the public site.

The founder

Omar Saleem, MD

Founder & Clinical Director

A South Asian physician who built Zinda because he saw the gap firsthand. The standard assessments did not always capture the context of South Asian patients. Zinda is the platform he wished existed when navigating these questions for his own community.

Follow the work

Long-form writing lives on the Zinda Substack: blog.zinda.health. The observatory on this site updates nightly from the research engine.