Capstone: design a Pakistani healthcare AI pilot end to end

Three ways to see it

1. 1

   Section one: problem and population. Name the facility, the patient group, and the clinical outcome you want to improve. Quantify the baseline if you can: how many cases, how often the current pathway fails, how much delay you are trying to remove. A vague problem statement guarantees a vague pilot. Write a problem statement that fits in three sentences and that the medical director can repeat back to you.

2. 2

   Section two: AI choice and evidence. Which AI approach fits and why. Cite at least two published references showing the approach works in a similar context. Note any Pakistani validation evidence. If no local validation exists, say how you will produce it in this pilot. State the false-positive and false-negative tolerance the clinical lead has agreed to.

3. 3

   Section four: data, consent, and governance. List the data you need by field. State the legal basis under PDPA 2023. Attach the patient consent form in Urdu and English. State retention and access policy. Name the IRB you will engage and the chair. Specify what happens if a patient withdraws consent mid pilot. Without this section the pilot cannot start.

The why-tree

Try this with Claude

Submission. Write the five sections as a single document, fewer than ten pages. Share it with one clinician, one IT lead, and one regulator-adjacent reader. Ask each for the weakest section. Revise once. Then commit to a sixteen-week timeline with a midpoint review. Pakistani healthcare AI does not need more demos. It needs a few well-designed pilots that land in clinical workflows and earn their place. Your capstone is the first step.