Web App

AI Triage

An AI pipeline for symptom analysis, test recommendation, evidence evaluation, and condition prediction.

Details

Role

Solo Full-Stack Developer

Project Type

Internship Project

Duration

1 Month

Status

Completed

Built With

JavaScript

Next.js

React.js

Chakra UI

HTML

CSS

Node.js

Express.js

MongoDB

Ollama

MedGemma 1.5 4B

What it is

An internship project that grew out of Pathify. It explores a structured pipeline that moves from reported symptoms to possible conditions, recommended laboratory tests, and evidence from the resulting test data.

Possible conditions are treated as guidance, not a diagnosis. The project is a controlled simulation of the workflow rather than a deployed clinical system.

Main features

AI Symptom Analysis

Processes reported symptoms as the starting point for the condition analysis flow.

Dataset-Driven Diagnosis

Combines curated medical datasets with the model so possible conditions stay grounded in known data.

Lab Test Recommendation

Suggests relevant laboratory tests based on the symptoms and possible conditions identified earlier in the pipeline.

Evidence-Based Evaluation

Uses entered test results to evaluate the earlier possibilities instead of producing a single answer directly from the symptoms.

Built with

Next.js and Chakra UI

Build the simulation screens for symptom input, test recommendations, result entry, and evidence review.

Node.js and Express.js

Provide the APIs that connect the different stages of the analysis pipeline.

MongoDB

Stores the datasets and session information used during the workflow.

Ollama

Runs the language model locally for the condition and evidence analysis steps.

MedGemma 1.5 4B

Provides the local model used for the AI-assisted analysis.

What came from it

The project explored a more structured approach to AI-assisted clinical decision support by combining medical datasets with model reasoning. Instead of asking a model for one direct answer, the pipeline breaks the process into symptoms, possible conditions, test recommendations, and evidence-based evaluation.

Screenshots

Simulation UI

Top section

Symptom input, suggested diagnostic tests, and laboratory result entry interface.

Video Demo

Next project

Pangita

A simpler way to discover local products and businesses.