Post Graduate Program in Generative AI for Business Applications

Generative AI is a subset of AI that uses Machine Learning models to learn the underlying patterns and structures in large volumes of training data. It then uses that understanding to create new data, such as images, text, videos, and more. This module provides a comprehensive overview of Generative AI models, how they evolved, and how to apply them e‰ectively to various business challenges.

(History of Generative AI, Generative AI vs Discriminative AI, Interacting with Generative AI, Overview of Hallucination, Business Applications of Generative AI, Overview of Pandas, Pandas Dataframes, Visual Analysis of Data)

Machine Learning (ML), a subset of Artificial Intelligence (AI), focuses on developing algorithms capable of learning patterns in data and making predictions without being explicitly programmed to do so. This module introduces participants to the notion of learning from data, what ML is and its types, and how to train and evaluate ML models for business applications.

(The Notion of Learning from Data, Introduction to ML, Types of ML Problem and Solution Space for ML, Exploratory Data Analysis, Training ML Models, Evaluating ML models)

Deep Learning (DL) is a branch of Machine Learning that leverages artificial neural networks (ANN) inspired by biological neurons and utilizes a collection of artificial neurons stacked and connected in layers to model complex data. This module dives deeper into the underlying functionality of neural networks and how to use common, open-source DL libraries Keras and Tensorflow to build neural networks to solve business problems.

(Overview of Neural Networks, Neural Network Architecture, Activation Functions, Gradient Descent, Traning a Neural Network, Backpropagation)

Embeddings allow us to numerically represent complex textual data. Transformers are neural network architectures that develop a context-aware understanding of data and have revolutionized the field of AI. This module provides a comprehensive overview of the role of embeddings in capturing meaning from text data, understanding the role of self-attention in the encoder component of transformers, and applying sentence transformers to enhance business applications.

(The Need for Embeddings, Introduction to Transformers, Encoder Component of a Transformer, Attention and Self-Attention, Sentence Embeddings with Sentence Transformers, Semantic Search Applications)

Industry - Finance

Summary - Analyze the data comprising stock news and prices and develop an AI-driven sentiment analysis system that will process and analyze news articles to gauge market sentiment to help financial analysts optimize investment strategies and improve client outcomes.

Tools & Concepts - Google Colab, Hugging Face, Transformers, Sentence Transformers

Decoder-only Transformers autoregressively generate text by predicting one word at a time based on previous words. This module will provide learners with a comprehensive view of the decoder component in transformers, the role of masking, cross-attention, and autoregressive generation, an understanding of how decoder-only transformer models process and generate text, and applying transformers to real-world business use cases.

(The Decoder Component of a Transformer, Masking, Cross Attention, Autoregressive Nature of the Decoder, Text Generation Applications)

Large Language Models (LLMs) are ML models that are pre-trained on large corpora of data and possess the ability to generate coherent and contextually relevant content. Prompt engineering is a process of iteratively deriving a specific set of instructions to help an LLM accomplish a specific task. This module introduces LLMs, explains their working, and covers practices to devise prompts to solve problems using LLMs effectively.

(Introduction to LLMs, Working of LLMs, Applications of LLMs, Introduction to Prompt Engineering, Strategies for Devising Prompts)

Retrieval augmented generation (RAG) combines the power of encoder and generative models to produce more informative and accurate outputs from an external knowledge source. This module will provide a thorough coverage of the importance of external knowledge sources in enhancing an LLM’s accuracy and contextual awareness, using vector databases to store and efficiently retrieve information from data, and evaluating the quality and relevance of the LLM-generated text.

(External Knowledge Sources, Data Chunking, Vector Databases, Retrieval-Augmented Generation (RAG), Evaluating RAG Systems)

Industry - Healthcare

Summary - Utilize sentence embeddings, vector databases, and Retrieval-Augmented Generation (RAG) to enhance information retrieval for a medical chatbot and provide accurate and context-aware responses., ensuring reliable and relevant medical guidance.

Tools & Concepts - Generative AI, Large Language Models, Prompt Engineering, Hugging Face, Retrieval Augmented Generation, Vector Databases.

Fine-tuning LLMs refers to the process of training a pre-trained large language model on domain-specific data to adapt it for specialized tasks, thereby improving its performance while retaining general language understanding. This module provides a comprehensive overview of the need for fine-tuning in adapting Large Language Models (LLMs) to specific business use cases, analyzing Parameter-Efficient Fine-Tuning (PEFT) techniques for optimizing model performance, and implementing QLoRA-based fine-tuning strategies to enhance LLM efficiency while minimizing computational costs.

(The Need for Fine-Tuning, Parameter-Efficient Fine-Tuning (PEFT), PEFT Techniques (Prefix Tuning, Prompt Tuning, QLoRA), QLoRA application and Implementation)

Agentic AI workflows involve methodologies for designing, automating, and managing the decision-making processes using AI agents to achieve specific goals. This module provides a comprehensive overview of LangChain, a versatile framework for integrating (LLMs) with external tools and services, covers different types of AI agents, and explores the architecture and design principles for building AI agents within LangChain.

(Introduction to AI Agents, Agentic AI Tools, AI Agents within LangChain, Agentic AI Workflows, Types of AI Agents)

Responsible AI involves developing AI systems that produce accurate outputs in a manner that is ethical, transparent, and fair, ensuring they benefit society while minimizing potential harm. This module delves into the critical aspects of AI ethics, provides a comprehensive overview of identifying and mitigating bias and risk in AI systems, covers the importance of ethical considerations in AI development, and an overview of the laws and regulations governing secure AI use.

(Identifying Bias and Risk in Human and AI Systems, Mitigating Bias and Risk in AI, Laws and Regulations for Responsible AI Use)

Industry - Legal

Summary - Utilize agentic AI workflows and LangChain to automate document analysis and summarization with an AI-powered legal document synthesizer by leveraging parameter-efficient fine-tuning techniques for efficient adaptation to legal language, mitigating bias, and minimizing risk in AI-generated content.

Tools & Concepts - Large Language Models, Prompt Engineering, Parameter-Efficient Fine Tuning, AI Agents, Responsible AI

This course will help you explore how to solve business problems by generating code using Generative AI tools, examine the capabilities of text-to-image and image-to-text GenAI tools like DallE through business use cases, and explore the speech recognition capabilities of audio-to-text GenAI tools like Whisper through business use cases.

• Image Captioning using GenAI
• Speech Recognition using GenAI
  

This course will provide you with an overview of the basic principles of MLOps and LLMOps, and help you explore how to deploy Generative AI solutions effectively using web applications, ensuring the scalability of the solutions to wider audiences for solving business problems.

• Overview of MLOps and LLMOps
• Deploying Generative AI Solutions using Web Apps