Agentic AI Training in Ahmedabad

Python basics and data structures

Functions, modules, and exception handling

Working with JSON and API responses

Environment setup and package management

Python programming exercises

Writing clear instructions for LLMs

System prompts and user prompts

Few-shot prompting

Output formatting and structured responses

Prompt testing and refinement

Handling ambiguous or incomplete inputs

What is an AI agent?

Chatbots, fixed workflows, and agents compared

Agent goals, instructions, and task boundaries

Planning and task decomposition

Agent execution cycles

Choosing appropriate agent use cases

Understanding tools and function calling

Defining tool inputs and outputs

Connecting agents to APIs

Validating tool arguments

Handling errors and failed requests

Permissions and controlled tool execution

Conversation context and state

Short-term and persistent memory concepts

Managing task history

Storing and retrieving relevant information

Managing context limits

Privacy considerations for stored data

Introduction to RAG

Document loading and text processing

Text chunking and embeddings

Vector databases and semantic search

Retrieving relevant information

Connecting retrieval to an AI agent

Evaluating answer relevance and grounding

Introduction to agent development frameworks

Defining agents and tools

Creating workflow steps

Managing agent state and execution

Error handling and debugging

Comparing framework-based and custom implementations

Introduction to multi-agent architectures

Specialised agent roles

Task delegation and handoffs

Sequential and parallel workflows

Shared context and coordination

Managing failures across agent workflows

When a single agent is preferable to multiple agents

Identifying suitable automation tasks

Connecting agents to external services

Designing repeatable workflows

Human approval and review steps

Handling exceptions and unexpected results

Monitoring workflow execution

Testing agent responses and actions

Measuring task completion and output quality

Tracing agent steps and tool calls

Handling hallucinations and unreliable outputs

Prompt injection and untrusted input risks

Guardrails and permission boundaries

Human oversight for sensitive actions

Preparing an agent application for deployment

Environment variables and API key protection

Logging and error monitoring

Latency and cost considerations

Reliability and performance testing

Maintaining and improving deployed workflows

Select a practical use case

Design the agent workflow

Integrate the required tools or knowledge sources

Test the agent using sample tasks

Document limitations and safety measures

Present the final application