Staff Engineer AI Agents

About the role

This is a rare chance to shape the future of how an entire industry operates — not in theory, but in production, at scale, touching real customers and physical assets every day. At Zuma, human and AI agents work side by side, and you'll help define what that collaboration looks like at its best.

Responsibilities

• Architect, build, and deploy production AI agents using modern agent frameworks (LangGraph, CrewAI, AutoGen, or equivalent), owning the full lifecycle from design to reliability in production.
• Define the technical patterns and standards for how agents are built across the engineering org — you will be setting the playbook others follow.
• Lead the rebuilding of core platform systems — including our onboarding/configuration system, integration framework, and AI performance analytics infrastructure.
• Collaborate directly with the VPE and product leadership to translate product vision into agent architecture, and make high-stakes technical trade-offs with confidence.
• Own agent reliability, observability, and continuous improvement — defining how we measure, monitor, and iterate on agent behavior in production.
• Work across the stack (backend services, LLM orchestration, integrations, data pipelines) to ship agents that are robust and scalable.
• Tame legacy code and lay down new foundations — patterns and architecture you create will be inherited by the engineers who come after you.
• Be a close partner to the product and operations teams, turning their domain needs into intelligent automated workflows without requiring domain expertise upfront.

Requirements

5+ years of software engineering experience with a strong backend or distributed systems foundation.

Demonstrated experience designing and shipping AI agents in production — not just prototypes. You've owned agent systems that real users depend on.

Hands-on experience with at least one modern agent framework such as LangGraph, CrewAI, AutoGen, Semantic Kernel, or a comparable orchestration layer.

Deep familiarity with LLM integration patterns — prompt engineering, tool/function calling, memory systems, retrieval-augmented generation (RAG), and agent evaluation.

Experience building reliable, observable agentic systems: tracing, error handling, fallback strategies, human-in-the-loop checkpoints, and graceful degradation.

Strong proficiency in Python and/or TypeScript — the languages our agents live in.

Nice to have

• Experience with multi-agent systems — coordination patterns, agent-to-agent delegation, and conflict resolution.
• Familiarity with vector databases and embedding strategies (Pinecone, Weaviate, pgvector, etc.).
• Prior experience at a startup or high-growth company; comfort shipping fast and iterating in production.
• Background in building self-serve platform or integration infrastructure.
• Experience with workflow automation platforms or business process orchestration.
• Experience with telephony integrations (Twilio or similar) and building voice-capable agents or chatbots across text and voice channels.
• Familiarity with speech-to-text, text-to-speech, or real-time audio streaming pipelines in production AI systems.
• Classical ML experience — supervised/unsupervised learning, feature engineering, model training and evaluation outside of LLM contexts.