Agent-Engineering-Lab

A hands-on repository for learning and building production-grade AI agents. Covers LLM fundamentals, RAG systems, tool calling, agent memory, planning, multi-agent architectures, MCP, observability, evaluation, security, and enterprise-scale agent platforms. Each module includes practical implementations, architecture diagrams etc.

Agentic AI Engineering Roadmap (Professional / Enterprise Level)

Phase 1: Foundation

01-Generative AI Applications

• Foundation
• Langchain with GenAI
• Customizing LLM and their outputs
• Building ChatBot with LLM
• Generative AI in Production

02-prompt-engineering

• Structured prompting
• XML prompting
• Chain of thought
• ReAct
• Self-reflection
• Constitutional prompting

---

Phase 2: Retrieval Systems

05-vector-search-engine

• Embeddings
• Similarity metrics
• ANN search

06-rag-engine

• Basic RAG
• Chunking
• Retrieval pipelines

07-advanced-rag

• Hybrid search
• Reranking
• Query rewriting
• Multi-hop retrieval

08-agentic-rag

• Planning-based retrieval
• Reflection loops
• Knowledge graph retrieval

09-long-term-memory-systems

• Episodic memory
• Semantic memory
• Memory compression
• Memory ranking

---

Phase 3: Tool Usage

10-tool-calling-engine

• OpenAI tools
• Structured outputs
• Function calling

11-api-integration-patterns

• REST
• GraphQL
• Event-driven tools

12-browser-agents

• Web navigation
• Playwright
• Browser automation

13-computer-use-agents

• Desktop automation
• Vision grounding
• UI interaction

---

Phase 4: Agent Architectures

14-single-agent-runtime

• Agent loops
• State machines
• Planning

15-planning-agent

• Task decomposition
• Execution planning

16-reflection-agents

• Self critique
• Repair loops

17-multi-agent-system

• Agent collaboration
• Delegation
• Consensus systems

18-human-in-the-loop-agents

• Approval workflows
• Escalation patterns

19-agent-orchestrator

• Workflow execution
• Agent routing

---

Phase 5: Frameworks

20-langgraph-playground

• Graph workflows
• State persistence

21-crewai-playground

• Crew orchestration

22-autogen-playground

• Conversational agents

23-openai-agents-sdk

• Production agent runtime

24-pydantic-ai

• Typed agents
• Structured execution

---

Phase 6: MCP Ecosystem

25-mcp-server

• Tool exposure
• Resource management

26-mcp-client

• Tool consumption

27-mcp-tool-registry

• Discovery patterns

28-mcp-agent-platform

• Enterprise MCP architecture

---

Phase 7: Production Engineering

29-agent-observability

• Tracing
• Metrics
• Logging

30-agent-evaluation

• Offline evaluation
• Online evaluation
• A/B testing

31-agent-security

• Prompt injection
• Tool abuse
• Data exfiltration

32-agent-reliability-engineering

• Retry strategies
• Fallback models
• Circuit breakers

33-cost-optimization

• Token budgeting
• Model routing
• Caching

---

Phase 8: Enterprise Systems

34-ai-workflow-engine

• Durable workflows
• Event-driven execution

35-enterprise-integrations

• CRM
• ERP
• Slack
• Jira
• GitHub

36-governance-compliance

• Audit trails
• SOC2
• GDPR
• HIPAA concepts

37-agent-platform

• Internal agent marketplace
• Shared memory
• Shared tools

38-enterprise-agent-cloud

• Multi-tenant architecture
• Agent hosting
• Isolation

---

Phase 9: Multimodal AI

39-vision-agents

• OCR
• Document understanding
• Image reasoning

40-audio-video-agents

• Voice assistants
• Realtime systems
• Video understanding

---

Phase 10: Capstone Projects

41-Enterprise Digital Workforce

Create a complete multi-agent organization.

42-research-agent

Build a Deep Research clone.

43-support-agent-platform

Enterprise customer support system.

44-autonomous-business-operator

Multi-agent business workflow automation.

45-Autonomous Procurement System

Procurement teams manually compare vendors and process requests.