Introduction to RAG & Agentic AI on Vertex AI
- Why
RAG is critical for enterprise AI
- RAG
vs fine-tuning
- Role
of agents in RAG systems
- Vertex
AI RAG ecosystem overview
Vertex AI Architecture for RAG Agents
- Vertex
AI generative models
- Embedding
models and vector stores
- Agent
orchestration components
- End-to-end
RAG system design
Data Ingestion & Knowledge Pipelines
- Document
ingestion patterns
- Chunking
strategies
- Metadata
and tagging
- Incremental
updates and re-indexing
Embeddings & Vector Search on Vertex AI
- Creating
embeddings
- Vector
indexes and similarity search
- Tuning
recall and precision
- Hybrid
search (keyword + vector) concepts
Retrieval Strategies & Query Understanding
- Query
rewriting and expansion
- Multi-query
retrieval
- Filtering
with metadata
- Re-ranking
strategies
Prompting for RAG Systems
- Prompt
templates for grounded answers
- Context
assembly strategies
- Citation-aware
prompting
- Reducing
hallucinations with grounding
Building RAG-Powered AI Agents
- Agent
roles and responsibilities
- Agent
+ RAG orchestration patterns
- Tool
calling with retrieved context
- Multi-step
agent workflows
Knowledge Validation & Answer Verification
- Source
attribution and citations
- Confidence
scoring concepts
- Cross-checking
retrieved content
- Handling
missing or conflicting information
Tool & System Integration
- Connecting
to enterprise systems
- APIs
and structured data sources
- Search,
databases, and document stores
- Secure
tool access for agents
Evaluation of RAG & Agent Systems
- Retrieval
quality metrics
- Answer
faithfulness and groundedness
- Automated
evaluation pipelines
- Human
review workflows
Performance & Cost Optimization
- Index
sizing and sharding
- Caching
strategies
- Token
optimization
- Cost
monitoring and limits
Security, Privacy & Data Governance
- Access
control for knowledge sources
- Handling
sensitive enterprise data
- Data
segregation and tenancy
- Compliance
and audit readiness
Monitoring & Observability
- Tracking
retrieval and generation metrics
- Detecting
drift in knowledge bases
- Logging
agent actions
- Alerting
and troubleshooting
Production Deployment Patterns
- CI/CD
for RAG pipelines
- Environment
separation (dev/test/prod)
- Rollback
and re-index strategies
- Versioning
prompts and indexes
Hands-on Exercises
Summary and Conclusion
