Lead AI Agent Engineer - Background Task Orchestration

Blue; Santa Clara (CA); On-site; USD 180,000 - 240,000; Full time

Job summary

Blue is building an integrated assistant that manages notifications, messages, and tasks across apps. The role spans live and background modes, with ownership of architecture, evaluation, and new capabilities.

Mid and senior levels are open, including a lead who will guide the background assistant's architecture and a small team. You will work across connectors, memory, safety, and cross-functional planning to deliver reliable, scalable assistant features.

Qualifications

• Shipped LLM agents or tool-using features to real users.
• Built evals and trust them beyond demos.
• Understand security risks of agents acting for a user, esp. prompt injection.
• Do the arithmetic early: latency, token budgets, cost per user.

Responsibilities

• Design and own the background assistant architecture: task planning, sub-agent handoffs, run and recover.
• Define evals, track quality, and monitor cost per user.
• Build and maintain connectors to APIs, MCP servers and private endpoints; enable new abilities like image generation and code PRs.
• Manage screen control and memory tools; ensure robust recall of user knowledge.
• Oversee safety and review: confirmations before risky actions, permissions, and agent changes.
• Collaborate with ML eng and product to plan implementation and delivery of assistant capabilities.

Skills

LLM agents

Eval metrics

Security risk awareness

Latency & budgets

Team leadership

Ambiguity tolerance

Tools

Android development

OAuth flows

MCP

Job description

Blue is building an integrated assistant that manages notifications, messages, and tasks across apps. The role spans live and background modes, with ownership of architecture, evaluation, and new capabilities.

Mid and senior levels are open, including a lead who will guide the background assistant's architecture and a small team. You will work across connectors, memory, safety, and cross-functional planning to deliver reliable, scalable assistant features.