Agentic AI Engineer

About Adzymic

Adzymic is an advertising technology company helping brands, agencies and media owners deliver creative and media campaigns across display, video, connected TV and digital out-of-home.

We are building AgenX, our agentic advertising platform, to connect creative generation, media discovery, campaign activation and insights through AI agents.

We are looking for an Agentic AI Engineer with hands-on experience in multi-agent orchestration, LangGraph, agent harness engineering and evaluations. You will work with our product and engineering teams to build reliable agents that complete real advertising tasks.

What you will do

• Build multi-agent architectures: Design agent responsibilities, routing, delegation, handoffs and shared state across creative, sales and campaign workflows.
• Develop workflows with LangGraph: Implement stateful workflows with conditional branching, parallel execution, checkpointing, persistence and human-in-the-loop approvals.
• Build agent harnesses: Develop the runtime and control layer around agents, including tool execution, permissions, retries, execution limits and failure recovery.
• Engineer context and memory: Manage how agents retrieve, prioritise and share information, including brand guidelines, inventory data, campaign context and conversation history.
• Develop evaluations: Build domain-specific datasets, scoring criteria and automated tests to assess task completion, tool-calling accuracy, output quality and failure handling.
• Monitor and improve performance: Use execution traces and production feedback to improve reliability, latency and cost. Run regression evaluations when models, prompts or tools change.
• Integrate advertising systems: Connect agents to creative tools, publisher systems and advertising platforms through APIs, MCP and relevant industry protocols.
• Ship production software: Build and maintain backend services with appropriate testing, deployment practices, documentation and monitoring.

What we are looking for

• At least 3 years of software engineering experience , or equivalent demonstrated ability, with hands-on experience building and deploying LLM or agentic applications.
• Strong Python skills and experience developing APIs and backend services.
• Practical experience with multi-agent orchestration , including agent coordination, state management and failure handling.
• Hands-on experience with LangGraph , including stateful graphs, conditional routing, persistence and human-in-the-loop workflows.
• Experience developing agent harnesses , including tool execution, context management, memory and execution controls.
• Experience with LLM and agent evaluations , including test datasets, scoring rubrics, automated evaluation pipelines and regression testing.
• Familiarity with observability tools such as LangSmith, Langfuse or equivalent , and the ability to diagnose failures from execution traces.
• Solid understanding of databases, asynchronous processing, authentication and reliable API integrations.
• Good judgement about when to use a single agent, multiple agents or deterministic application logic.
• Ability to collaborate with product and business teams, explain technical trade-offs and take features through to production.

Good to have

• Experience with MCP, agent-to-agent communication or other agent interoperability approaches.
• Familiarity with AdCP, AAMP, Prebid, OpenRTB or programmatic advertising.
• Experience with retrieval-augmented generation, vector databases and structured knowledge retrieval.
• Experience with multimodal models, creative generation or automated creative quality checks.
• Familiarity with cloud infrastructure, containers and CI/CD.
• Experience building SaaS applications with customer data isolation and role-based permissions.

Prior advertising industry experience is helpful but not required.

What success looks like

• Agents complete defined advertising tasks reliably, with measurable improvements in task success, speed and cost.
• Workflows recover from failures and maintain clear approval boundaries for consequential actions.
• Evaluation and monitoring systems detect regressions and guide ongoing improvements.
• Integrations and code are maintainable, documented and reusable across AgenX.

Why join us

• You will build AI products for real advertising workflows, working closely with product, commercial and operations teams. You will have meaningful ownership of your work and help shape how brands, agencies and media owners use AI agents.