AI Engineer at Lumion

Lumion•South Jordan, UT

3d•Onsite

About The Position

Lumion is seeking an AI Engineer to own the systems behind Lumion's AI in production. This role is critical as AI is the core product at Lumion, powering conversational agents that interact with prospective students, answer their questions, and guide them through the enrollment process. The systems handle sensitive information including money, student records, and regulated communication, requiring a high degree of reliability and accuracy. The AI Engineer will be responsible for building and maintaining conversational agents, the retrieval layer for accurate, school-specific answers, multi-model infrastructure for cost and latency optimization, voice capabilities for realistic phone conversations, and robust reliability measures like evaluations, guardrails, and monitoring.

Requirements

Shipped production AI systems that users relied on.
Experience dealing with hallucination, runaway costs, and latency limits in AI systems.
Strong fundamentals in data modeling, API design, and system architecture.
Ability to debug production incidents under pressure.
Understanding of product sense and why requirements exist, not just how to build them.
Fluency with agentic coding tools like Claude Code, Codex, or Cursor.
Ability to recognize incorrect AI output and understand long-term consequences of quick fixes.
Must be authorized to work in the US. We do not sponsor work visas.

Nice To Haves

Experience in product engineering with fast iteration cycles.
Experience working in an AI-native engineering culture.

Responsibilities

Own the systems behind Lumion's AI in production.
Build and manage conversational agents that interact with people via text, email, and live phone calls.
Develop the logic for agent outreach timing, messaging, and human handoff.
Own the retrieval layer to ensure accurate answers based on specific school data.
Manage model selection, cost, and latency for different AI tasks.
Develop voice capabilities to make AI phone calls indistinguishable from human calls.
Implement evaluations, guardrails, and monitoring to ensure AI system reliability.