https://github.com/codewitheshayoutube/acemed_ai
https://github.com/codewitheshayoutube/acemed_ai
Last synced: 2 months ago
JSON representation
- Host: GitHub
- URL: https://github.com/codewitheshayoutube/acemed_ai
- Owner: codewithEshaYoutube
- License: apache-2.0
- Created: 2025-03-16T11:22:28.000Z (over 1 year ago)
- Default Branch: main
- Last Pushed: 2025-05-08T09:08:04.000Z (about 1 year ago)
- Last Synced: 2025-05-08T10:24:07.884Z (about 1 year ago)
- Language: Python
- Size: 19.2 MB
- Stars: 0
- Watchers: 1
- Forks: 0
- Open Issues: 0
-
Metadata Files:
- Readme: README.md
- License: LICENSE
- Codeowners: .github/CODEOWNERS
Awesome Lists containing this project
README
# AceMed AI 🇵🇰
AI-powered MDCAT preparation platform tailored to PTB/Federal board syllabus using LLaMA-based fine-tuned models.
---
## 🖥️ Landing Page
### AceMed AI — Pakistan’s AI-Powered MDCAT Preparation Platform
**Crack the MDCAT — Score Higher with AI Precision**
AceMed AI is trained on exact FSC Federal, PTB, and provincial board books, offering precise, personalized preparation powered by advanced AI.
👉 [Start Free Trial](https://acemed-ai.streamlit.app/)
👉 [Learn More](https://acemed-ai.streamlit.app/)
---
## 🌐 Navigation
- Home
- About Us
- Features
- Pricing
- FAQs
- **Login | Register**
---
## 🎯 Key Highlights
### Why AceMed AI?
AceMed AI merges technology with education to create a personalized and efficient MDCAT prep experience. It replaces costly coaching centers with smart, adaptive tools, empowering students to:
- Learn at their own pace
- Focus on weak areas
- Practice with real-time AI feedback
**Mission:** To make high-quality MDCAT prep accessible, smart, and personalized through the power of AI.
---
AceMed AI is Pakistan’s first AI-powered MDCAT preparation assistant, trained on PTB and Federal Board books. It provides:
- ✍️ Accurate, syllabus-based answers
- 📊 Performance analytics dashboard
- 🤖 Chatbot interface for MDCAT Q&A
- 📚 Curated MCQ banks
- 🔍 Step-by-step numerical solvers
- 🔄 Adaptive learning and feedback loops
---
## 🛠️ Technology Stack
| Layer | Tools & Frameworks |
|----------------|------------------------------------------------|
| **Frontend** | Streamlit, HTML, TailwindCSS, React.js |
| **Backend** | FastAPI, LangChain, Python, HuggingFace |
| **Database** | MongoDB, Redis |
| **AI Models** | LLaMA, LoRA Fine-Tuning, Transformers |
| **Infra** | Google Colab Pro+, GitHub, Vercel |
---
## 🔬 AI Model Fine-Tuning
We leverage the **LLaMA model**, fine-tuned using **LoRA (Low-Rank Adaptation)** to adapt the base language model to the **MDCAT domain**. This enables precise, context-aware responses aligned with FSc and PTB syllabi.
👉 [Open Fine-Tuning Notebook in Colab](https://colab.research.google.com/drive/19h9IH47HhXx30C2gfd7Kr6GzJwB6-2-Y)
### 📘 Fine-Tuning Overview
| Component | Details |
|--------------------------|-------------------------------------------------------------------------|
| **Base Model** | Meta LLaMA (7B) |
| **Fine-Tuning Method** | LoRA (Low-Rank Adaptation) |
| **Data Used** | Curated PTB + Federal Board textbook content + MDCAT MCQs |
| **Framework** | 🤗 Hugging Face Transformers + PEFT |
| **Notebook** | Google Colab for rapid iteration |
| **Training Objective** | SFT (Supervised Fine-Tuning) on syllabus-aligned Q&A |
| **Epochs** | 3–5 (adaptive based on validation loss) |
| **Optimization** | AdamW optimizer, 5e-5 learning rate |
| **LoRA Ranks** | r=8, alpha=16 |
| **Hardware Used** | Google Colab Pro+ (A100 GPU) |
---
## 🔄 Development Roadmap
AceMed AI follows the **Agile Software Development Life Cycle (SDLC)** for rapid iteration, scalability, and user-focused features.
### 📅 Phases
1. **Research & Requirement Analysis**
- Aligning features with PMC syllabus & student feedback
2. **Data Collection & Preprocessing**
- Structuring PTB/Federal board content and MCQs
- Cleaning & labeling data for model training
3. **AI Model Development**
- Fine-tuning transformer-based models
- Developing step-by-step numerical solvers
4. **System Development & Integration**
- Backend infrastructure + chatbot interface
- Interactive dashboard for performance analytics
5. **Testing & Deployment**
- Unit & integration testing
- Beta deployment for real-world usage
6. **User Feedback & Iteration**
- Feedback from students & educators
- Feature refinement & model accuracy tuning
---
## 💸 Pricing
| Plan | Price | Features |
|-----------------|---------------|---------------------------------------------------------------------------|
| 🎓 **Free** | Rs 0/month | Limited AI question generation, Basic analytics |
| 🥇 **Gold** | Rs 2500/month | Unlimited AI-generated questions, Detailed performance analytics, Support |
| 👑 **Platinum** | Rs 5000/month | All Gold features, 1-on-1 mentoring, Exclusive MCQ banks |
---
## ❓ FAQs
**Q: How is AceMed AI different from ChatGPT?**
A: It's trained specifically on MDCAT syllabus (Federal/PTB), ensuring relevant, accurate answers.
**Q: Can AceMed AI improve my marks?**
A: Yes, through adaptive learning and targeted practice.
**Q: Is past paper practice included?**
A: Yes, along with textbook references and explanations.
**Q: How accurate are the answers?**
A: 95%+ based on internal testing. Manual reviews ongoing for edge cases.
---
## 🤝 Contribution Guidelines
AceMed AI is open-source and welcomes contributions:
1. Fork the repository
2. Create a feature branch
3. Commit your changes
4. Open a pull request
---
## 📜 License & Links
- **License**: MIT License
- **Website**: [AceMedAI.com](https://acemedai.com)
- **Contact**: support@acemed.ai |
---