AI Engineer, Agent Analytics & Optimization

Conviva•Foster City, CA

22h•$180,000 - $250,000•Hybrid

About The Position

Conviva is seeking an AI Engineer to build systems that enable AI agents to understand, query, and act on Conviva's behavioral data. This is a hands-on builder role focused on designing and shipping production-grade agentic systems, from tool interfaces to deployment and iteration. The role involves creating MCP servers, agentic workflows, and AI connectors that integrate Conviva's platform with external systems. Success will be measured by the ability to build and ship reliable, scalable systems, own the full lifecycle of these systems, and collaborate effectively with product and customers. This role is distinct from Product Builder roles, focusing on the engineering execution rather than roadmap definition.

Requirements

3–7+ years of software engineering experience, with production systems shipped end-to-end.
Strong Python backend fundamentals; comfortable owning services from design through deployment (fullstack a plus).
Hands-on experience building and operating LLM-powered agents in production — with concrete stories of what broke, how you fixed it, and what you'd do differently.
Practical familiarity with agentic patterns: tool use, structured outputs, prompt chaining, multi-agent coordination, and knowing when a deterministic workflow is the right call.
Experience designing/building MCP servers or equivalent tool-use / AI connector integrations.
Have used evals to validate and debug agent behavior.
Comfortable with cloud deployment (GCP/AWS), containerized services, and async Python.
Strong product instinct — thinks about what users actually need and pushes back when a spec doesn't add up.

Nice To Haves

Experience with agentic frameworks (LangGraph, Anthropic SDK, Pydantic-AI, CrewAI, or similar).
Agent observability/tracing (Langfuse, LangSmith, OpenTelemetry, or homegrown).
Background in digital experience analytics, observability, or APM.
Familiarity with agent evaluation or LLM-ops tooling.
Published writing, talks, or open-source work at the intersection of engineering and AI.

Responsibilities

Design and build MCP servers that expose Conviva's data and analytics as discoverable, reliable tools for AI agents, with proper auth, scoping, and access controls.
Architect and ship agentic workflows: multi-step reasoning, tool use, orchestration patterns, and human-in-the-loop where needed.
Develop AI connectors integrating Conviva's intelligence with external agents, clients, and downstream systems.
Own the full lifecycle — API/tool design, deployment, monitoring, production hardening.
Use evals to validate agent behavior before shipping, catch regressions, and debug production divergence.
Build observability and tracing to make agent behavior inspectable and auditable.
Tune for latency, cost, and reliability as models and data drift.
Partner with Product Builders, design, and customers to ship working software (not prototypes).
Build internal SDKs, reusable patterns, and shared context that raise the floor for the whole team.
Stay close to the fast-moving MCP/agent ecosystem and bring back what's relevant.