← All work

08 · Personal & Academic Project

Medical Image Analyzer

Designing two reliable explanations from one clinical image

  • Python
  • GPT-4o
  • Prompt Engineering
  • Streamlit
  • API Security

Business value

One image produced both a structured professional report and a clearer patient-facing explanation.

01 · Business problem

Clinicians and patients need different outputs from the same X-ray, while raw multimodal prompts can produce inconsistent structure and language.

02 · My approach

I treated prompt design like interface design: separate audiences, explicit roles, structured sections, and secure handling.

03 · What I built

A clinical-report prompt, a plain-language explanation prompt, a Streamlit workflow, and secure API configuration.

What changed my thinking

The corrections, tradeoffsand decisions behind it.

  1. 01

    One answer could not serve two audiences

    Clinical precision and plain-language understanding required separate prompt paths.

  2. 02

    Structure improved reliability

    Explicit output sections reduced variation more than a generic request for a better answer.

  3. 03

    Prompting was product design

    The challenge was building a repeatable experience around model capability.