Engineer

Kargo creates powerful moments of connection between brands and consumers to build businesses. Every day, our dynamic teams work to radically raise the bar on what agentic AI, CTV, eCommerce, social, and mobile can do to deliver unique ad experiences across the world’s most premium platforms. Taking a Creative Science approach to all we do, we continuously innovate solutions that outperform industry benchmarks and client expectations. Kargo is growing rapidly and currently has offices in NYC, Chicago, LA, Dallas, Sydney, Auckland, London and Waterford, Ireland. 

Who We Hire

Techies who want to build the future. Creatives who want to design it better. Communicators to win business. Collaborators to build it. Data pros who turn numbers into insights. Product builders who turn ideas into innovations. Anyone eager to be on a team that doesn’t stop to ask what’s next, because they’re already building it. 

The Opportunity

Karlo is Kargo's agentic ad-buying platform, where customers use conversational AI agents to build campaigns, creatives, targeting strategies, and analytics. As an AI Platform Engineer on the Karlo Platform team, you'll own the shared infrastructure every agent runs on — observability, model evaluation, the LangGraph runtime, checkpointing and context memory management, shared services, and the developer experience tying it all together. Your goal is to make the platform invisible to the agent teams building on it: so reliable, scalable, and well-documented that they can focus entirely on their agents.

The Daily To-Do

• Own and harden the core LangGraph-based agent hosting infrastructure so every agent team can build and deploy with confidence — no undocumented behaviors, no surprise failures
• Design and deliver a shared evaluation framework that lets agent teams validate correctness, measure quality, and catch regressions before they reach customers
• Ship production-ready shared services — authentication, billing, and integration layers including Salesforce and CXP — that agent teams consume cleanly instead of rebuilding
• Own checkpointing and agent context memory — state persistence, history compaction, and retrieval of relevant prior context across long-running, multi-turn conversations
• Instrument the platform end-to-end: every agent, every session, every failure, with tracing across LLM calls plus latency, cost, and failure-mode visibility
• Build the SDKs, documentation, and onboarding tooling that cut a new agent team's ramp time from weeks to days
• Partner closely with agent teams to translate their needs into platform capabilities, and measurably reduce the infrastructure overhead they carry today