Cognizant GenAI Externship – Projects & Reflections

Welcome to my GitHub repository for the Cognizant Generative AI Externship. This repo contains a collection of work completed during the externship, including hands-on Python exercises, fine-tuned generative models, capstone projects, and written reflections.

🔍 Overview

Throughout the externship, I:

• Completed structured Python programming exercises and standalone projects
• Built and fine-tuned 4 generative AI models using PyTorch, Hugging Face, and Cohere’s API
• Explored prompt engineering and model parameter tuning
• Reflected on model performance and limitations through written reports

---

📁 Repository Structure

Assignments/

Introductory Python exercises containing multiple small problems per file (a1.py to a6.py). Topics include:

• Lists, dictionaries, loops, and conditionals
• Function writing and basic algorithmic problem-solving

Projects/

Longer, standalone Python programs designed to reinforce problem-solving. Examples:

• calculator.py: Basic command-line calculator
• groceryTracker.py: Simple grocery list tracker
• passwordChecker.py: Password validation tool

Capstone/

Higher-level projects where I made more design decisions and built full Python programs with increased complexity and structure.

---

🤖 GenAI Model Notebooks

The four main GenAI model projects are located inside their respective folders under Projects/. Each one explores a different model or technique:

gan_assignment.ipynb (in root)

Used BigGAN to generate images from random latent vectors. Explored how varying input noise affects image output.

transformer_template.ipynb (in Projects/ExploringTransformers/)

Used GPT-2 to generate text from prompts. Experimented with prompt types, temperature, and max token settings.

bert.ipynb & fineTuningBERT.pdf (in Projects/BertFineTuning/)

Fine-tuned a BERT model on a sentiment analysis task using public Twitter data. Evaluated performance and learned debugging techniques.

creativeApp.ipynb (in Projects/BertFineTuning/)

Built a text generation script using Cohere’s GPT-3.5 API. Designed and tested different prompt strategies and parameter settings to evaluate generation quality.

---

📝 Reflections

Each major modeling project is paired with a brief written reflection (PDF) covering:

• Model purpose and architecture
• Experiment process
• Observations and takeaways

Located in project folders (e.g., fineTuningBERT.pdf, Cognizant_transformers.pdf).

---

💻 Technologies Used

• Python
• PyTorch
• Hugging Face Transformers
• Cohere API
• Jupyter / Google Colab

---