Writing LinkedIn posts consistently is harder than it sounds. Most people either skip posting or end up with the same generic "Excited to share..." opener. This tool generates multiple unique versions of a post from a single input — different angles, different tones, ready to use.
| Feature | Details |
|---|---|
| 📌 Post Types | Achievement, Learning, Project Launch, Internship, Hackathon, Reflection |
| 🎭 Tone Modes | Professional, Conversational, Storytelling, Motivational |
| 🔁 Multiple Versions | Up to 3 unique posts per generation, each with a different CTA |
| 📊 Post Stats | Word and character count shown for each post |
| ⬇️ Download | Save any post as a .txt file |
| 🎨 UI | Dark theme, fully responsive |
Never starts with "Excited to" or "Thrilled to".
User Input (topic / achievement)
│
▼
┌─────────────────┐
│ Prompt Builder │ ← Post type + tone + variation logic
└────────┬────────┘
▼
┌─────────────────┐
│ Groq LLM API │ ← Llama 3 70B inference
└────────┬────────┘
▼
┌─────────────────┐
│ Post Formatter │ ← 3 unique versions with hashtags + CTA
└────────┬────────┘
▼
┌─────────────────┐
│ UI / Output │ ← Preview, copy, download
└─────────────────┘
linkedin-post-generator/
├── public/ # Firebase-hosted frontend
│ ├── index.html
│ ├── app.js
│ └── style.css
├── app.py # Streamlit app (local)
├── generator.py # Post generation logic
├── requirements.txt
└── screenshots/
Input: Completed Microsoft Azure internship via AICTE Elevate. Built a fake image detector using MobileNetV2.
Output (Conversational):
Six weeks ago, I didn't know what transfer learning meant. Today, I shipped a fake image detector using MobileNetV2 — trained on the CIFAKE dataset, deployed as a Streamlit app, and presented as my capstone for the Microsoft Elevate Azure Internship. It's not perfect. But it works. And more importantly — I built it.