Agent Engineer – San Francisco, CA

AHU technology; San Francisco, Northern (CA, KY); Hybrid; USD 160,000 - 200,000; Full time

Benefits offered by this job

Equity

Job summary

AHU technology is seeking an Agent Engineer in San Francisco to ship production-grade agent and workflow layers for a cloud-based creative canvas. You’ll connect image, video, audio, text, 3D, and tool nodes into reusable production workflows, orchestrating multiple AI models and providers with robust memory and routing systems.

You will implement planning, tool use, memory, retrieval, context management, and reliable workflow execution.

Qualifications

• 3+ years shipping agent, LLM, backend, infrastructure, or AI-product systems in production.
• Strong Python and/or TypeScript backend experience with databases, queues, observability, and production reliability.
• Hands-on experience with agent/runtime design: planning, tool use, memory, retrieval, context management, or workflow execution.
• Experience orchestrating multiple model providers, including routing, retries, fallbacks, latency, cost, and failure handling.
• Ability to build practical eval loops, regression testing, human review, A/B benchmarks, or quality metrics.
• Strong product judgment for creative workflows, where the system needs to do useful work on a canvas - not just chat.

Responsibilities

• Build the agent runtime for a node-based creative canvas, including planning, tool use, memory, retrieval, context, and workflow execution.
• Orchestrate multiple AI models and providers across image, video, audio, text, and tool workflows.
• Design routing, retries, fallbacks, latency control, cost awareness, and failure handling for generation systems.
• Build workflow memory that preserves project context, prior outputs, style choices, assets, and reusable templates.
• Create quality loops through evals, regression tests, human review, benchmarks, and creator-facing metrics.
• Improve production reliability through queues, cancellation, progress states, monitoring, and graceful recovery.

Skills

Python/TypeScript

Agent runtime design

Model orchestration

Observability

Evaluation loops

Product judgment

Tools

Databases

Message queues

Observability tools

Production reliability

Job description

We are hiring an Agent Engineer for a confidential AI creative platform. This person will build the agent and workflow layer behind a cloud-based creative canvas where users connect image, video, audio, text, 3D, and tool nodes into reusable production workflows. The ideal candidate has shipped real production AI or backend systems, understands agent runtime behavior, and can build reliable model orchestration beyond simple prompt demos.
Work arrangement: Onsite
Salary: $160K-$200K + equity

WHAT YOU WILL OWN

• Build the agent runtime for a node-based creative canvas, including planning, tool use, memory, retrieval, context, and workflow execution.
• Orchestrate multiple AI models and providers across image, video, audio, text, and tool workflows.
• Design routing, retries, fallbacks, latency control, cost awareness, and failure handling for generation systems.
• Build workflow memory that preserves project context, prior outputs, style choices, assets, and reusable templates.
• Create quality loops through evals, regression tests, human review, benchmarks, and creator-facing metrics.
• Improve production reliability through queues, cancellation, progress states, monitoring, and graceful recovery.

WHAT WE ARE LOOKING FOR

• 3+ years shipping agent, LLM, backend, infrastructure, or AI-product systems in production.
• Strong Python and/or TypeScript backend experience with databases, queues, observability, and production reliability.
• Hands-on experience with agent/runtime design: planning, tool use, memory, retrieval, context management, or workflow execution.
• Experience orchestrating multiple model providers, including routing, retries, fallbacks, latency, cost, and failure handling.
• Ability to build practical eval loops, regression testing, human review, A/B benchmarks, or quality metrics.
• Strong product judgment for creative workflows, where the system needs to do useful work on a canvas - not just chat.
• Must be able to work on-site in San Francisco.