AI Engineer

Job Summary We are seeking an AI / Applications Engineer to build the agents and applications that put our data to work. You will design, build, and evaluate AI agents that automate real operational workflows, and you will build the interfaces through which people interact with them. This is an applied engineering role rather than a research one: success is measured by whether agents produce correct, trusted answers in production, and whether people actually use what you ship. We care far more about your judgment in building reliable LLM systems than about which specific platform you have used. Our warehouse runs on Snowflake, but the patterns transfer, and we expect the right candidate to come up to speed on our stack quickly. Responsibilities • Design, build, and operationalize AI agents that automate operational workflows across the business. • Develop and register the tools and skills that agents invoke — SQL and function tools, API tools, search services — and integrate them into agent frameworks and orchestration layers. • Own agent evaluation: define what a correct answer looks like before building, build evaluation sets, and measure accuracy, hallucination, latency, and inference cost over time. • Monitor agents in production and act on what the measurements show, including regression detection when prompts, models, or underlying data change. • Build web interfaces and data applications that put data, models, and agents in front of users, including internal tools and agent front-ends. • Build retrieval pipelines — chunking, embedding, and indexing strategies — to power semantic search and grounding. • Contribute to the context layer that agents depend on, structuring and enriching trusted data for retrieval-augmented generation and text-to-SQL, in partnership with the analytics engineering function. • Manage tool access and permissions so that agents reach only the data they should. • Partner with business stakeholders to identify workflows worth automating, scope them realistically, and set expectations about what these systems can and cannot do. • Maintain clear documentation of agent design, tool definitions, prompts, and evaluation results.