GenerativeAI with Claude
By Kiran Kumar
Build AI Agents and Agentic Systems with Claude. A 30-hour intermediate program covering the Claude model family, prompt engineering, tool use, RAG, MCP, agent design patterns and the Claude Agent SDK, with a hands-on capstone project.
About the Program
Generative AI with Claude: Building AI Agents and Agentic Systems is a 30-hour program (10 modules of 3 hours each, or 5 days of 6 hours). It takes you from LLM and prompt-engineering fundamentals through the building blocks of agents — tool use, RAG and MCP — into full agentic systems, evaluation and deployment.
About 40% concepts and demos, 60% hands-on labs. Every module ends with a lab, and the program closes with a capstone project where you build a complete agent of your own.
Orientation with the Trainer
Learn more about the generative AI course directly from the trainer himself. He introduced the core concepts during this weekly webinar.
Weekly Webinar Introduction
In this session, the trainer walks through the course flow, the core concepts of Generative AI with Claude, and what you'll be able to build by the end of the program.
Watch the full webinarCurriculum
Generative AI and Claude Fundamentals
Part A: Foundations · 3 hrs
- How LLMs work: tokens, context windows, sampling, temperature
- The Claude model family and how to pick a model for cost, speed and capability
- Anthropic's approach to safety and reliability, and what it means for builders
- Ways to use Claude: the Claude app, Claude Code, API and Console
- Lab: Set up the Console, create API keys, make your first Messages API call in Python and TypeScript
- Lab: Experiment with system prompts, multi-turn conversations and streaming
Prompt Engineering and Context Engineering
Part A: Foundations · 3 hrs
- Principles of clear prompting: role, task, constraints, examples
- Structuring prompts with XML tags, few-shot examples and output formats
- Reasoning techniques: extended thinking, step-by-step decomposition, self-checking
- Context engineering: what to include, what to leave out, long-context best practices
- Prompt caching to reduce cost and latency
- Lab: Build and iterate on prompts for classification, extraction and summarization
Structured Outputs, Tool Use and Function Calling
Part B: Building Blocks of Agents · 3 hrs
- Why tools turn an LLM into an agent
- Defining tools: names, descriptions, JSON schemas
- The tool-use loop: request, tool_use, execute, tool_result
- Parallel tool calls, forcing tool choice, and error handling
- Structured JSON outputs and validation with Pydantic or Zod
- Lab: Build a weather, calculator and database-lookup assistant with a hand-written tool loop
Retrieval-Augmented Generation (RAG) and Knowledge Grounding
Part B: Building Blocks of Agents · 3 hrs
- Why RAG: hallucinations, freshness, private data
- Embeddings, chunking strategies and vector stores
- Retrieval quality: hybrid search, reranking, contextual retrieval
- Citations and grounded answers with Claude
- Agentic RAG: letting the agent decide when and what to retrieve
- Lab: Build a document Q&A assistant over a company knowledge base with citations
Model Context Protocol (MCP)
Part B: Building Blocks of Agents · 3 hrs
- The problem MCP solves and its architecture: hosts, clients, servers
- MCP primitives: tools, resources, prompts
- Using existing MCP servers (filesystem, GitHub, databases, SaaS apps)
- Building a custom MCP server in Python or TypeScript
- Security: authentication, permissions, prompt injection through tools
- Lab: Build and connect an MCP server that exposes an internal API to Claude
Agent Architecture and Design Patterns
Part C: Agentic Systems · 3 hrs
- Workflows vs. agents: when to use which, and why to start simple
- Core patterns: prompt chaining, routing, parallelization, orchestrator-workers, evaluator-optimizer
- The agent loop: plan, act, observe, reflect
- Memory: short-term context, long-term stores, summarization and compaction
- Designing good tools for agents (the agent-computer interface)
- Lab: Implement two patterns from scratch (routing and evaluator-optimizer)
Building Agents with the Claude Agent SDK
Part C: Agentic Systems · 3 hrs
- Overview of the Agent SDK and its built-in capabilities
- Built-in tools: file operations, shell, web search and fetch
- Custom tools and MCP integration in the SDK
- Permissions, hooks and guardrails
- Sessions, streaming and long-running tasks
- Lab: Build a research agent that searches the web, reads sources and writes a cited report
Multi-Agent Systems, Skills and Claude Code
Part C: Agentic Systems · 3 hrs
- When multi-agent helps: parallel research, specialization, context isolation
- Subagents: delegation, handoffs, and aggregating results
- Agent Skills: packaging reusable instructions and scripts
- Claude Code for engineers: agentic coding, CLAUDE.md, slash commands, hooks, MCP
- Hosted and managed agent options on the Claude platform
- Lab: Build a coordinator with 2-3 specialist subagents (security, testing and docs agents)
Evaluation, Safety, Observability and Deployment
Part D: Production and Capstone · 3 hrs
- Evaluating agents: test sets, LLM-as-judge, trajectory and outcome metrics
- Safety and security: prompt injection, least-privilege tools, human-in-the-loop approvals, sandboxing
- Reliability: retries, rate limits, timeouts, fallbacks, idempotency
- Cost and latency optimization: model selection, prompt caching, batch API
- Observability: logging, tracing and monitoring agent runs
- Deployment options: serverless, containers, cloud platforms
- Lab: Create an eval suite for your agent and improve it based on failures
Capstone Project and Wrap-up
Part D: Production and Capstone · 3 hrs
- Project briefing and architecture review, followed by the capstone build
- Option 1: Customer support agent with RAG, ticketing tools and escalation
- Option 2: Data analyst agent that queries databases and produces reports
- Option 3: Software engineering agent for issue triage, code changes and tests
- Option 4: Business operations agent integrating email, calendar and docs over MCP
- Demos, peer feedback, best practices checklist, assessment and feedback
What You'll Learn
- Explain how LLMs work and choose the right Claude model for cost, speed and capability
- Design effective prompts with structured outputs, reasoning techniques and prompt caching
- Build the tool-use loop that turns an LLM into an agent, with validated JSON outputs
- Create grounded answers with RAG, citations and agentic retrieval
- Build and connect custom MCP servers that expose internal APIs to Claude
- Apply agent design patterns and build agents with the Claude Agent SDK
- Evaluate, secure, observe and deploy agents in production
- Ship a complete capstone agent for support, analytics, engineering or operations
Who Should Attend
- Developers and engineers with basic Python or TypeScript
- QA professionals and testers moving into AI-powered automation
- Technical professionals familiar with REST APIs and JSON
- Builders ready to move from prompting to full Agentic AI systems
Prerequisites & Setup
- Python 3.10+ or Node.js 18+, a code editor and Git
- An Anthropic Console account with API credits
- Basic JSON, REST and command-line knowledge
Course Format
Live, Hands-On Sessions
30 hours across 10 modules (3 hours each) or 5 days (6 hours each). About 40% concepts and demos, 60% hands-on labs, with a lab in every module.
Schedule
Free demo sessions: 16th, 17th & 19th November. Paid program starts 30th November.
Pricing & Enrollment
Enroll in the Generative AI with Claude program.
Generative AI with Claude
FAQ
Do I need an AI or coding background?
The program is intermediate level: basic Python or TypeScript and REST API familiarity are assumed. Module 1 covers the Generative AI and Claude fundamentals, then it builds up step by step with hands-on labs.
Will we use real AI tools in the sessions?
Yes. The sessions cover popular AI tools and show how the concepts apply across them, with practical demonstrations.
When do the sessions start?
Free demo sessions are on 16th, 17th and 19th November. The paid program starts on 30th November.
What will the pricing or membership options be?
You can register for the session using the registration link above. Pricing and membership details will be announced with the schedule.