2-Day Generative AI Workshop: Industrial Insights & Hands-On for Final Year CS Students

This workshop is designed to provide 7th-semester Computer Science engineering students

with practical, industry-relevant knowledge and hands-on experience in Generative AI,
preparing them for potential careers in this rapidly evolving field. The focus will be on real-
world applications, tools, and best practices.

Day 1: Foundations of Generative AI & Large Language Models (LLMs)

Morning Session (9:00 AM - 12:30 PM): Introduction to GenAI & Industrial Landscape

 9:00 AM - 9:30 AM: The Rise of Generative AI

o What is Generative AI? Differentiating it from traditional AI/ML.

o Brief history and evolution: From rule-based systems to deep learning and
generative models.

o Key types of Generative Models: LLMs, Diffusion Models, GANs, VAEs (brief
overview).

o Why GenAI is a game-changer for industries (e.g., content creation, software

development, design, customer service).

 9:30 AM - 10:30 AM: Core Concepts & Architectures

o LLMs Deep Dive: Transformers architecture (Encoder-Decoder vs. Decoder-

only for LLMs).

o Attention Mechanism, Self-Attention, Positional Encoding (intuitive

explanation, not mathematical derivation).

o Pre-training vs. Fine-tuning vs. Prompting.

o Key LLM models: GPT series, LLaMA, Gemini, Claude (brief comparison and
their industrial applications).

 10:30 AM - 10:45 AM: Break

 10:45 AM - 11:45 AM: Industrial Landscape & Career Paths in GenAI

o Current trends and market demand for GenAI skills.

o Typical job roles: Prompt Engineer, ML Engineer (GenAI focus), Data Scientist
(GenAI), AI Product Manager, AI Researcher.

o Skills required: Python, ML frameworks (PyTorch/TensorFlow), cloud

platforms, understanding of model architectures, MLOps, communication.

o Case studies of companies leveraging GenAI (e.g., GitHub Copilot,

Midjourney, Jasper AI).
  11:45 AM - 12:30 PM: Ethical Considerations & Responsible AI

o Bias and fairness in GenAI models.

o Hallucinations and factual accuracy.

o Data privacy and security.

o Intellectual property and copyright issues.

o The importance of Responsible AI development and deployment.

Lunch Break (12:30 PM - 1:30 PM)

Afternoon Session (1:30 PM - 4:30 PM): Hands-On with Large Language Models

 1:30 PM - 2:15 PM: Introduction to Prompt Engineering

o The art and science of communicating with LLMs.

o Basic prompt structures: Zero-shot, Few-shot, Chain-of-Thought.

o Techniques for effective prompting: Clear instructions, role-playing,

constraints, examples.

o Iterative prompt refinement.

 2:15 PM - 3:45 PM: Hands-On Lab: Interacting with an LLM API

o Setup: Quick setup of a Python environment (Jupyter/Colab) and API key

(e.g., Gemini API, Hugging Face transformers library with a suitable open-
source model).

o Task 1: Text Generation:

 Generating creative content (stories, poems).

 Generating marketing copy or product descriptions.

 Controlling output length and style.

o Task 2: Summarization:

 Summarizing long articles or documents.

 Extracting key information.

o Task 3: Question Answering & Chatbots:

 Building a simple Q&A system.

 Simulating a basic conversational agent.

o Task 4: Code Generation/Assistance (if time permits):

  Generating simple code snippets.

 Debugging assistance.

 3:45 PM - 4:15 PM: Debugging & Troubleshooting LLM Outputs

o Common issues: Repetitive text, irrelevant responses, factual errors.

o Strategies for improving output: Adjusting temperature, top-p, max tokens,

refining prompts.

 4:15 PM - 4:30 PM: Q&A and Day 1 Wrap-up

Day 2: Image Generation, Advanced Applications & Career Readiness

Morning Session (9:00 AM - 12:30 PM): Generative AI for Images & Other Modalities

 9:00 AM - 10:00 AM: Introduction to Diffusion Models

o Intuitive explanation of how Diffusion Models work (forward and reverse

diffusion).

o Key concepts: Latent space, noise scheduling, U-Net architecture (high-level).

o Comparison with GANs (advantages of Diffusion Models).

o Popular models: Stable Diffusion, DALL-E, Midjourney.

 10:00 AM - 11:30 AM: Hands-On Lab: Image Generation

o Setup: Using a readily available API or library (e.g., Hugging Face diffusers
with a pre-trained model, or a web-based tool like Stable Diffusion
Playground).

o Task 1: Text-to-Image Generation:

 Generating images from textual prompts.

 Exploring prompt variations for different styles (e.g., "photorealistic,"

"oil painting," "cyberpunk").

o Task 2: Image-to-Image / Inpainting (if API supports):

 Modifying existing images.

 Filling in missing parts of an image.

o Task 3: ControlNet (conceptual overview):

 How to control image generation with sketches, poses, etc.

(demonstration if possible, otherwise theoretical).

 11:30 AM - 12:15 PM: Beyond Text & Images: Other Generative Modalities
 o Audio Generation: Music, speech synthesis (e.g., text-to-speech, voice
cloning).

o Video Generation: Text-to-video, image-to-video.

o 3D Model Generation: Generating meshes and textures.

o Code Generation (Advanced): More complex code structures, unit test

generation.

 12:15 PM - 12:30 PM: Challenges in Multi-Modal GenAI

o Data requirements, computational costs, coherence across modalities.

Lunch Break (12:30 PM - 1:30 PM)

Afternoon Session (1:30 PM - 4:30 PM): Real-World Applications, MLOps & Career
Readiness

 1:30 PM - 2:30 PM: Industrial Use Cases & Case Studies

o Software Development: Automated code reviews, test case generation,

documentation.

o Marketing & Advertising: Personalized ad copy, creative asset generation.

o Healthcare: Drug discovery, medical image synthesis.

o Finance: Fraud detection (synthetic data generation), report automation.

o Gaming & Entertainment: Procedural content generation, character design.

o Education: Personalized learning content, intelligent tutoring systems.

 2:30 PM - 3:30 PM: MLOps for Generative AI

o The unique challenges of deploying and managing GenAI models.

o Model monitoring (drift, bias detection).

o Version control for models and data.

o Continuous Integration/Continuous Deployment (CI/CD) for GenAI.

o Scalability and cost optimization for inference.

o Fine-tuning pipelines in production.

 3:30 PM - 4:00 PM: Mini-Project/Challenge & Ideation Session

o Students work in small groups to brainstorm and outline a GenAI project idea
based on a real-world problem.
 o Present their ideas (e.g., a GenAI-powered content assistant for a specific
industry, a tool for generating synthetic data, an AI art generator with specific
constraints).

o Discussion on feasibility, required data, and potential challenges.

 4:00 PM - 4:15 PM: Next Steps & Career Guidance

o Resources for continued learning: Online courses, communities, research

papers.

o Building a portfolio: Showcasing GenAI projects.

o Networking in the AI community.

o Interview preparation tips for GenAI roles.

 4:15 PM - 4:30 PM: Open Q&A and Workshop Conclusion

Prerequisites for Students:

 Basic understanding of Python programming.

 Familiarity with fundamental Machine Learning concepts (e.g., supervised vs.

unsupervised learning, neural networks at a high level).

 Laptop with internet access.

Tools & Platforms (Suggested):

 Google Colab or Jupyter Notebooks for hands-on labs.

 Access to a Generative AI API (e.g., Google Gemini API, OpenAI API, Hugging Face
API/models).

 Relevant Python libraries: transformers, diffusers, torch/tensorflow.

This workshop aims to be highly interactive, with a balance of theoretical understanding and
practical application, ensuring students gain valuable insights and skills for their future
careers in Generative AI.