Persona view
Explore how different roles engage with each layer of the Velocity Blueprint. Each cell represents a specialized role responsible for a specific focus within a layer.
Business/Product Product vision, priorities, and ROI | Architecture Technical design and standards | Build/Dev Development and implementation | Reliability Operations and incident response | Governance Security, compliance, and policy | |
|---|---|---|---|---|---|
| Agentic core | |||||
| Software core | |||||
| Data core | |||||
| Foundations |
Understanding the view
Layers
- Agentic core
- Software core
- Data core
- Foundations
Concerns
- Business/Product: Product vision, priorities, and ROI
- Architecture: Technical design and standards
- Build/Dev: Development and implementation
- Reliability: Operations and incident response
- Governance: Security, compliance, and policy
Agent engineer
Build/Dev · Agentic core
Builds and maintains AI agents, prompts, and tools.
Responsibilities
- • Develop and test agent prompts
- • Implement agent tools and integrations
- • Debug agent behavior
- • Optimize agent performance
Key metrics
Primary components (9)
Instruments agent code with trace spans and investigates behavioral anomalies
Agentic core › Evaluation frameworkWrites evaluation test cases and benchmarks agent performance across scenarios
Agentic core › Prompt debuggingDebugs prompt failures, analyzes token usage, and optimizes prompt chains
Agentic core › Prompt managementAuthors, versions, and iterates on agent prompts and system instructions
Agentic core › Knowledge groundingImplements RAG retrieval logic, manages knowledge base connectors, and tunes relevance
Agentic core › Tool frameworkBuilds and registers agent tools, defines tool schemas, and tests tool integrations
Agentic core › MCP serversBuilds and tests MCP server integrations for tool access and data retrieval
Agentic core › Agent development kit (ADK)Uses the agent SDK daily for scaffolding, testing, and deploying agents
Agentic core › Orchestration frameworksImplements agent orchestration logic using frameworks like LangGraph or CrewAI
Secondary components (6)
Integrates session state to maintain conversational context across interactions
Agentic core › Long-term memoryImplements memory persistence for agents that learn from prior interactions
Software core › Code agentsUses AI coding agents for rapid prototyping and automated code generation
Software core › Test agentsBuilds automated test agents for regression and behavioral validation
Software core › AI pair programmingLeverages AI pair programming for real-time coding assistance
Software core › Dev MCP toolsConfigures development-time MCP tools for local agent testing