AI Engineer, Agent Analytics & Optimization (DPI)

About this role

Build production-grade AI agent systems that enable autonomous tools to understand and act on Conviva's behavioral analytics. Design MCP servers, orchestrate multi-step agentic workflows, and ship integrations at scale—working end-to-end from tool design through deployment and observability.

What you'll do

• Design and build MCP servers exposing Conviva's data as discoverable tools for AI agents with proper authentication and access controls
• Architect multi-step agentic workflows including reasoning patterns, tool orchestration, and human-in-the-loop interactions
• Develop AI connectors integrating Conviva's intelligence with external agents, clients, and downstream systems
• Own full lifecycle of systems: API design, deployment, monitoring, and iteration based on production usage
• Work closely with product and customers to translate direction into reliable systems
• Ensure observability and reliability of agent systems running at scale

What they're looking for

• Backend/systems engineering with production deployment experience
• AI/ML systems design (agents, tool use, orchestration patterns)
• API design and integration
• Authentication and access control systems
• Observability and monitoring practices
• Multi-step workflow design and state management
• Python or similar systems language
• Experience with LLM tooling frameworks (e.g., Claude SDK, model context protocol)

Benefits

• Work on emerging AI agent infrastructure at a profitable, industry-leading company
• Foster City, CA location
• Build production systems solving real problems at scale, not research or prototyping
• End-to-end ownership from design through runtime and customer impact
• Collaboration with established enterprise customers (Disney, NBC, NFL)

Likely interview questions

• Describe your experience designing APIs or tool interfaces for external systems or agents. How did you ensure discoverability and reliability?
• Walk us through a multi-step workflow or orchestration system you've built. How did you handle state, retries, and error cases?