Offered by the Technocraft Centre for Applied AI, IIT Bombay

This module introduces the conceptual building blocks for working with AI systems. Learners set up Python, familiarise themselves with core machine learning concepts, and understand how data flows through modern pipelines. Short primers on deep learning and NLP (embeddings) establish a shared vocabulary. The module also covers how modern models represent language, why embeddings matter, and how these representations power downstream tasks. By the end, learners can explain how meaning is encoded numerically and why this foundation is essential for LLM applications. Topics Covered: - Python for Generative AI - Advanced Python for Generative AI Applications - Introduction to Machine Learning - Introduction to Deep Learning - NLU and NLP: Word Embeddings

This module moves from fundamentals to the core mechanics of modern language models. Learners unpack attention mechanisms and transformer architecture before applying these concepts to practical prompting. The module demonstrates how LLMs handle both classification and labelling tasks, as well as open-ended text generation and summarisation, highlighting what changes, what to measure, and how to steer outputs eectively. Topics Covered: - Attention Mechanism and Transformers - Large Language Models and Prompt Engineering - LLMs for Classification and Text Labelling - LLMs for Text Generation and Summarisation

In this module, learners will explore the evolution from CNNs and VAEs/GANs to diffusion-based foundation models, and how text pairs with images/audio for real tasks. They will learn multimodal prompting patterns and how to build and evaluate simple multimodal apps. Topics Covered: • Evolution of Multimodal Models - from CNNs to VAEs, GANs, Transformers, and Modern Diffusion-based Foundation Models • Applications: Vision-Language Models, Audio-Visual Reasoning, Robotics, etc. • Multimodal Prompting & Interaction • Building & Evaluating Multimodal Applications

This module focusses on designing production-ready workflows that ground models in enterprise data. Topics include Retrieval-Augmented Generation (RAG), application blueprints, and selective fine-tuning for accuracy and tone. The module concludes with clear decision-making guides on when to use RAG, when to fine-tune, and how to evaluate performance. Topics Covered: - Retrieval-Augmented Generation - Advanced RAG: Optimisation and Multi-Modal RAG - Designing and Building LLM Workflows - Fine-Tuning Custom LLMs

This module translates single-prompt prototypes into multi-step systems that plan, use tools, and manage tasks collaboratively. Agent capabilities and patterns are introduced first, followed by building single-agent flows with memory and tool integration. The module ends with multi-agent collaboration and principled evaluation to support trust in agent behaviour. Topics Covered: - Introduction to AI Agents - Building Single-Agent Systems and Agentic RAG - Multi-Agent Collaboration and Agent Evaluation

The final module addresses deploying applications responsibly and ensuring their ongoing reliability. The concepts covered include CI/CD for LLM apps, deployment patterns, and live monitoring for latency, cost, drift, and quality. The module also covers vulnerabilities, data governance, and operational controls so that releases are secure, compliant, and observable from day one. Topics Covered: - CI/CD and DevOps for LLM Applications - LLMOps: Generative AI Application Deployment - LLMOps: LLM Application Performance Monitoring - Mitigating LLM Vulnerabilities and Ensuring Data Governance