AI-First Engineering Manager, Agent Runtime

Box Com; Redwood City (CA); Hybrid; USD 180,000 - 260,000; Full time

Job summary

Box is building the infrastructure for AI agents across its product ecosystem. We are seeking an engineering leader to run the agent runtime platform, overseeing a six-engineer hybrid team and collaborating with product and engineering leaders to deploy agent-based solutions at scale.

You will champion an eval-driven culture, own architecture decisions, and present progress to VP-level stakeholders, while actively contributing code when needed.

Qualifications

• 7+ years of software or ML engineering experience.
• 2+ years of engineering management experience leading hybrid teams.
• Hands-on experience with AI agents or LLM-based systems and agent architectures.
• Experience deploying customer-facing agents in production at scale.

Responsibilities

• Lead and grow a six-person hybrid team of software engineers and ML engineers.
• Own the onboarding of internal tenant teams onto the agent runtime.
• Drive the technical architecture of the agent runtime service, including the context retrieval layer.
• Champion an eval-driven development culture with evaluation frameworks and metrics.
• Manage and coordinate the team’s on-call rotation and incident response.
• Present weekly plans, progress, and demos to VP-level leadership.
• Contribute directly to technical solutions and code as a player-coach.

Skills

Software engineering

ML engineering

Engineering management

AI agents

Distributed systems

Cross-functional collab

Tools

LangGraph

Dspy

AutoGen

Job description

Box is building the infrastructure for AI agents across its product ecosystem. We are seeking an engineering leader to run the agent runtime platform, overseeing a six-engineer hybrid team and collaborating with product and engineering leaders to deploy agent-based solutions at scale.

You will champion an eval-driven culture, own architecture decisions, and present progress to VP-level stakeholders, while actively contributing code when needed.