CourseBot

A specialized AI model trained to assist students with course-related questions, leveraging advanced fine-tuning techniques and optimized model architecture.

Key Achievements

• Model Optimization:

  • Optimized Mistral-7B model using QLoRA techniques and 4-bit quantization
  • Achieved 22.2% improvement in question-answering accuracy for course content
  • Implemented efficient inference pipeline with minimal computational overhead

• Data Engineering:

  • Synthesized comprehensive training corpus using Claude API and prompt chaining
  • Generated and validated 400+ high-quality Q&A pairs for fine-tuning
  • Ensured diverse coverage of course materials and concepts

• Technical Innovation:

  • Constructed robust context injection framework
  • Developed custom evaluation suite for response quality
  • Implemented reliable knowledge retention mechanisms
  • Ensured consistent course-specific responses

Setup

Install required packages: pip install transformers datasets accelerate bitsandbytes torch peft trl PyPDF2 scikit-learn pandas

Project Structure

• Base_model_inference.ipynb: Base model inference implementation
• Data Preprocessing.ipynb: Data preparation pipeline
• Training.ipynb: Model training and fine-tuning
• finetuned_model_inference.ipynb: Fine-tuned model inference

Requirements

• Python 3.11+
• PyTorch
• Transformers
• Other dependencies listed in installation command

Usage

1. Run data preprocessing notebook
2. Execute training pipeline
3. Use inference notebooks for predictions