AI Agents Roadmap

Learn to design, build and ship AI agents in 2026

Learn the Pre-requisites

LLM Fundamentals
  • Transformer Models and LLMs
  • Model Mechanis
  • Tokenization
  • Context Windows
  • Token Based Pricing
  • Generation Controls
  • Temperature
  • Frequency Penalty
  • Presence Penalty
  • Top-p
  • Stopping Criteria
  • Open Weight Models
  • Closed Weight Models
  • Model Families and Licences
  • Streamed vs Unstreamed Responses
  • Reasoning vs Standard Models
  • Fine-tuning vs Prompt Engineering
Understand the Basics of RAG
  • Understanding RAG
  • Embeddings and Vector Search
  • Vector Databases
  • Chroma, FAISS, Pinecone
  • RAG Optimization
  • Hybrid Search
  • Context Compression
  • Re-ranking
  • Query Expansion
Basic Backend Development
  • Git and Terminal Usage
  • REST API Knowledge
  • Backend Beginner Roadmap
  • Git and GitHub Roadmap
  • API Design Roadmap

AI Agents 101

What are AI Agents?
  • Definition and Core Concepts
  • Agent vs Chatbot
  • Autonomous vs Assisted Agents
  • Types of Agents
Agent Loop
Perception / User Input
↓
Reason and Plan
↓
Acting / Tool Invocation
↓
Observation & Reflection
↓
Back to Reasoning
Example Usecases
  • Personal assistant
  • Code generation
  • Data analysis
  • Web Scraping / Crawling
  • NPC / Game AI
  • Research Assistant
  • Content Creation
  • Customer Support

Prompt Engineering

What is Prompt Engineering
  • Be specific in what you want
  • Provide additional context
  • Use relevant technical terms
  • Use Examples in your Prompt
  • Iterate and Test your Prompts
  • Specify Length, format etc
Writing Good Prompts
  • Chain-of-Thought Prompting
  • Few-shot Learning
  • Role Prompting
  • Instruction Prompting
  • Reflective Prompting
  • Prompt Chaining
Prompt Engineering Roadmap
  • Beginner to Advanced
  • Industry Best Practices
  • Tool-specific Prompting

Tools / Actions

Tool Definition
  • Name and Description
  • Input / Output Schema
  • Error Handling
  • Usage Examples
  • Tool Metadata
  • Tool Versioning
Examples of Tools
Common Tools
  • Web Search
  • Code Execution / REPL
  • Database Queries
  • API Requests
  • Email / Slack / SMS
  • File System Access
  • Calendar Integration
  • Payment Processing

Agent Memory

What is Agent Memory?
  • Short Term Memory
  • Long Term Memory
  • Within Prompt
  • Vector DB / SQL / Custom
  • Episodic vs Semantic Memory
Maintaining Memory
Memory Strategies
  • RAG and Vector Databases
  • User Profile Storage
  • Summarization / Compression
  • Forgetting / Aging Strategies
  • Memory Consolidation
  • Memory Retrieval Optimization

Agent Architectures

ReAct (Reason + Act)
  • Paper Implementation
  • Reasoning Steps
  • Action Execution
  • Observation Integration
Model Context Protocol (MCP)
Other Architecture Patterns
  • Chain of Thought (CoT)
  • RAG Agent
  • Planner Executor
  • DAG Agents
  • Tree-of-Thought
  • Hierarchical Agents
  • Collaborative Agents
  • Reflexion Agents

Building Agents

Manual (from scratch)
  • Direct LLM API calls
  • Implementing the agent loop
  • Parsing model output
  • Error & Rate-limit handling
  • State Management
  • Tool Integration
LLM Native "Function Calling"
  • OpenAI Functions Calling
  • Gemini Function Calling
  • OpenAI Assistant API
  • Anthropic Tool Use
  • Cohere Tool Use
Building Using Frameworks
  • Langchain
  • LlamaIndex
  • Haystack
  • AutoGen
  • CrewAI
  • Smol Depot
  • Anthropic Tool Use
  • OpenAI Assistants API

Evaluation and Testing

Metrics to Track
  • Success Rate
  • Response Time
  • Accuracy
  • User Satisfaction
  • Tool Usage Efficiency
  • Error Rate
  • Task Completion Time
Testing Methods
  • Unit Testing for Individual Tools
  • Integration Testing for Flows
  • Human in the Loop Evaluation
  • LangSmith
  • Ragas
  • DeepEval
Frameworks
  • Structured logging & tracing
  • Debugging and Monitoring
  • Observability Tools
  • LangSmith
  • Helicone
  • LangFuse
  • openllmetry

Security & Ethics

Prompt Injection / Jailbreaks
  • Types of Attacks
  • Defense Strategies
  • Testing for Vulnerabilities
  • Prompt Hardening
  • Input Sanitization
Tool sandboxing / Permissioning
  • Sandbox Environments
  • Permission Systems
  • Access Control
  • Rate Limiting
  • Audit Logging
Data Privacy + PII Redaction
  • PII Detection
  • Data Redaction
  • Privacy-by-Design Principles
  • Compliance Requirements
  • Data Minimization
Bias & Toxicity Guardrails
  • Content Moderation
  • Bias Detection
  • Ethical Guidelines
  • Output Filtering
  • Stakeholder Input
Safety + Red Team Testing
  • Red Team Strategies
  • Safety Evaluations
  • Incident Response
  • Threat Modeling
  • Continuous Testing

Related Roadmaps

AI Engineer Roadmap
AI and Data Scientist Roadmap
MLOps Roadmap
AI Red Teaming Roadmap
Prompt Engineering Roadmap

AI Agents Roadmap - A community effort to help developers grow in their career

© 2026 roadmap.sh