We are seeking an Agentic AI Orchestration Architect for a short-term engagement supporting our client’s Visual Production team. The goal is to build durable orchestration that connects the team’s existing agents, automations, applications, and user-owned workflows without unnecessary rebuilding. Each stage will retain its own workflow and human review process. The orchestration layer will maintain shared job state, manage routing and handoffs, recognize approvals and completion events, propagate changes downstream, and provide end-to-end visibility. The architecture should use stable interfaces so agents, models, and tools can evolve independently. We expect value to be delivered incrementally. Working capabilities should be released into daily production use as they are ready, rather than held for a single launch at the end.
PROJECT SCOPE:
— Confirm current Visual Production workflows, systems, assets, handoffs, approvals, exceptions, open decisions, and bottlenecks; identify rules requiring team resolution
— Identify the authoritative systems for job state, assets, approvals, and final outputs, and document the target workflow and a milestone plan before building the first release
— Design a canonical work-order schema and durable state model covering job ID, assets, references, asset/version lineage, routing, status, ownership, approvals, decisions, and audit history
— Build the orchestration layer to track job and stage state, route work within and between user workflows, collect outputs, trigger downstream work, and identify stalled or blocked jobs
— Support stage-level human review, revision, and manual completion events, advancing work only when required approvals or completion signals are received, including physical production steps such as studio shoots and print proofs
— Build conditional routing based on job state and asset readiness, including bypassing stages when required assets already exist, and safely propagate upstream changes to affected downstream work
— Connect existing agents, automations, and systems through APIs, webhooks, MCP, and other appropriate interfaces; automate file and status handoffs across Google Workspace, Box, Figma, shared file repositories, and related tools
— Build for reliability and deliver in incremental production milestones, releasing working capabilities into daily use as they are ready; refine with team feedback, document the system and train the team as each milestone ships
DELIVERABLES:
— Workflow architecture: Current- and target-state specification, architecture diagram, work-order schema, state model, and milestone plan for sequencing capabilities into incremental releases
— Working orchestration layer: Delivered in production milestones, beginning with a first release on live jobs as soon as dependencies allow and expanding to conditional routing, stage-level approvals, manual completion events, downstream change propagation, and exception handling
— Integrations and visibility: Agreed agents and systems connected through stable interfaces, with job status, ownership, asset/version lineage, approval, and decision history visible across handoffs
— Handoff: Documentation, runbook, training, and prioritized backlog sufficient for the client to operate, troubleshoot, and extend the system
EXPERIENCE:
— 5+ years building production software, creative technology, automation, integration, or workflow systems, including recent hands-on agentic AI experience
— Demonstrated experience deploying production agentic or multi-agent workflows and explaining their architecture, reliability model, and operating approach
— Workflow and state-machine design skills, including conditional routing, shared state, human-in-the-loop processes, events, retries, and idempotency
— Hands-on Python or TypeScript development with REST APIs, webhooks, OAuth, structured outputs, and JSON schemas
— Experience with agent/tool interoperability patterns, including MCP or equivalent approaches
— Production experience developing OpenAI agents; comparable experience with other production agent platforms also valued
— Hands-on integration experience with Google Workspace APIs (Drive, Slides), Box, and shared file-server or DAM environments; Figma API experience preferred
— Ability to design model- and framework-agnostic systems using stable interfaces rather than tightly coupling workflows to one agent, model, or vendor
— Security fundamentals and experience navigating enterprise IT and security review for AI and workflow integrations, including least-privilege access, credential management, third-party systems, and confidential assets
— Experience translating nontechnical team workflows into usable systems and delivering fixed-scope engagements incrementally through production milestones, documentation, training, and handoff
NICE TO HAVE
— Creative operations or production-tracking experience across image generation, retouching, 3D/rendering, DAM, or other large-asset workflows
— Workflow automation platforms such as n8n, Make, Zapier, or equivalent
— Agent interoperability approaches and orchestration frameworks such as A2A, LangGraph, Google ADK, or similar.
— Comfort learning automation interfaces for creative tools such as Photoshop, 3ds Max, Figma Weave, and ComfyUI; scripting, Google Apps Script, or agent observability experience is a plus
#LI-Onsite
