This repository contains two end-to-end Generative AI projects built using Hugging Face Transformers and Python. These projects demonstrate practical implementation of modern NLP models for text generation and question answering tasks.

📌 Project 1: AI Storyteller – Text Generation using Transformer Models

🎯 Objective

To build a Generative AI system that creates coherent and creative stories based on user prompts using pre-trained transformer-based language models.

---

🧠 Model Used

• Pretrained Transformer-based Language Model (e.g., GPT-style architecture)
• Loaded using Hugging Face pipeline("text-generation")

---

⚙️ How It Works

1. Load a pretrained text generation model.

2. Accept a prompt from the user.

3. Generate a continuation of the story.

4. Control creativity using:

   • max_length
   • temperature
   • top_k
   • top_p

---

🧾 Example

Input Prompt:

Once upon a time in a futuristic city

Generated Output:

Once upon a time in a futuristic city, robots and humans lived side by side...

---

📊 Analysis

✅ Strengths

• Generates creative and fluent text.
• Handles open-ended prompts well.
• Produces coherent short stories.

⚠️ Limitations

• May sometimes generate repetitive text.
• Longer outputs may lose coherence.
• Factually incorrect content possible (hallucinations).

🎯 Learning Outcome

• Understanding generative language models.
• Working with decoding strategies.
• Controlling creativity vs determinism.

---

📌 Project 2: Ask Me Anything – Question Answering with BERT

🎯 Objective

To build an extractive question-answering system that returns factual answers from a given context paragraph.

---

🧠 Model Used

• bert-base-cased-squad2
• Loaded using Hugging Face pipeline("question-answering")
• Based on BERT architecture trained on SQuAD2 dataset

---

⚙️ How It Works

1. Load the Question Answering pipeline.
2. Provide a context paragraph.
3. Accept one or more questions.
4. The model extracts the most relevant answer span from the context.

---

🧾 Example

Context

Artificial Intelligence is a branch of computer science.
Machine Learning is a subset of AI.
Python is widely used in AI development.

Questions & Model Output

Question	Model Answer	Confidence
What is Artificial Intelligence?	A branch of computer science	0.98
What is Machine Learning?	A subset of AI	0.99
Which language is used in AI?	Python	0.97

---

📊 Analysis

✅ Strengths

• High accuracy for factual, direct questions.
• Extracts precise answer spans.
• Confidence score helps evaluate reliability.

⚠️ Limitations

• Cannot answer if information is not in context.
• Struggles with reasoning-based or inferential questions.
• Sensitive to poorly structured context.

🎯 Learning Outcome

• Understanding extractive QA systems.
• Working with contextual embeddings.
• Evaluating model confidence and performance.

---

🔍 Comparative Analysis: Generative vs Extractive AI

Feature	AI Storyteller	Ask Me Anything
Type	Generative AI	Extractive QA
Output Style	Creates new text	Extracts from context
Creativity	High	None
Accuracy	Variable	High (if answer exists in context)
Hallucination Risk	High	Very Low
Use Case	Chatbots, content creation	Search systems, knowledge assistants

---

🛠 Tech Stack

• Python
• Transformers (Hugging Face)
• PyTorch
• Jupyter Notebook

---

💡 Real-World Applications

AI Storyteller

• Creative writing assistants
• Content generation
• Marketing copy generation
• Story-based games

Ask Me Anything

• Customer support automation
• Document search systems
• FAQ bots
• Internal knowledge base systems