This repository showcases advanced Generative AI projects that combine the power of transformer architectures, diffusion models, and LLM pipelines to generate images, text, and intelligent responses. Each project reflects state-of-the-art methods in modern AI, as explored in the IBM AI Engineering Professional Certificate coursework and independent research.
| Project | Description | Tools & Models |
|---|---|---|
| Aircraft Damage Captioning | Classifies aircraft damage and generates natural language captions from images. | VGG16, BLIP, Keras, Hugging Face |
| Diffusion Models | Implements image generation using denoising diffusion probabilistic models. | NumPy, Matplotlib, Forward/Reverse Sampling |
| Text Generation with Transformers | Generates creative and coherent text using Transformer-based language models. | GPT2, Hugging Face Transformers |
| Advanced Transformers | Custom Transformer architecture from scratch with positional encoding and masked attention. | TensorFlow/Keras |
- Transformers (Hugging Face, BLIP, GPT2)
- TensorFlow/Keras
- LangChain (for RAG)
- Matplotlib, NumPy, PIL
- Diffusers, Datasets, Scikit-learn
- Image Captioning with BLIP using pretrained transformer encoders + decoders
- Diffusion model training loop from scratch with noise schedules
- Text generation with GPT2 (sampling, top-k, and temperature control)
- Retrieval-Augmented Generation (RAG) architecture with external context and vector databases (coming soon)
pip install transformers tensorflow datasets diffusers langchain pillow matplotlib numpyThis project is licensed under the MIT License.