I gave Gemini a job at a hospital

I built Medscribe AI, a full-stack AI medical scribe for doctor-patient conversations. It records or accepts an audio upload, cleans the audio with denoising, normalization, and resampling, sends it to Groq Whisper for transcription, and then uses Gemini to generate a structured clinical summary, doctor feedback, and a patient follow-up email.

This project combines Next.js, FastAPI, TypeScript, Python, Docker, Hugging Face Spaces, Vercel, Clerk authentication, Upstash Redis rate limiting, DeepFilterNet audio cleanup, Pydantic, and Zod into one end-to-end AI workflow.

In this video, I walk through the architecture, the transcription pipeline, the AI analysis flow, the deployment setup, and the full-stack engineering decisions behind the project.

Tech stack: Next.js, React, TypeScript, FastAPI, Python, Docker, Hugging Face Spaces, Vercel, Groq Whisper, Gemini, DeepFilterNet, Pydantic, Zod, Clerk, Upstash Redis, Nodemailer, GitHub Actions

What MedScribe AI does:

Records or uploads clinical audio
Cleans noisy audio before transcription
Converts speech to text with AI
Generates a structured clinical summary
Produces doctor feedback and follow-up actions
Drafts a patient-friendly follow-up email
Runs as a deployed full-stack AI system

This is a project demo and engineering walkthrough, not medical advice or a medical device.

#Nextjs #FastAPI #TypeScript #Python #Docker #HuggingFace #Vercel #AI #MachineLearning #SpeechToText #Groq #Gemini #FullStack #WebDevelopment #MedicalAI #AIMedicalScribe #SoftwareEngineering

========

🩻 App Link:
https://medscribe-ai-lilac.vercel.app/dashboard

🤖 Source Code:
https://github.com/zeyad-shaban/medscribe-ai-frontend
https://github.com/zeyad-shaban/medscribe-ai-backend

Видео I gave Gemini a job at a hospital канала ZeyadCode