AI Engineer

Every day, millions of shipments move around the world on data that nobody is looking at in real time. Orders, ASNs, invoices, and tracking events flow between 3PLs, forwarders, brands, and retailers, then get flattened into a status that tells you something went wrong only after it already has.SwiftB2B exists to change that. We are an AI operating system for global supply chains. One platform where 3PLs, freight forwarders, brands, retailers, and manufacturers can see every item, shipment, and exception as it happens. The platform does not stop at flagging a delay. It works out what the delay costs, weighs the options, and starts the fix.We built it on 15 years of SwiftAnt’s work in supply chain and integration technology, and on a team that has spent careers in logistics, retail, and enterprise software. We are a Databricks partner and our data platform runs on the lakehouse.A demo that passes and a system that answers are different things - we know that anyone can show a chatbot answering “which shipments are late?” on a good day. The hard part is making it answer from governed metrics only, refusing when it should, and proving a pass rate before a customer ever sees it. This role is a lot about that. You will make “ask the platform anything” true rather than aspirational, and build the agents that detect an exception, decide what it costs, and escalate rather than guess.What you will doOwn the natural-language query architecture on Databricks: Genie, the semantic layer as ground truth, and the guardrail layer that makes sure natural language resolves against governed metrics and never against free-form SQL on raw tables.Build and run the evaluation harness. Fifty questions per industry pack, drawn from the KPI catalogue, with expected answers, at or above ninety per cent before anything is demonstrated or deployed. Publish the pass rate, good or bad.Try to break your own guardrail, and fix what you find. When a question fails, work out whether the fix belongs in the prompt or in the metric definition. Usually it is the definition.Build agent workflows on the platform: alerting, narrative generation, exception triage, corrective actions. Agents that escalate when uncertain, with a decision trace someone can re-examine later.Put a number on every query: tokens, latency, cost per answer. Keep a technology watch on models and decide when a rule beats a model.Teach analysts and customers what the system can and cannot answer, so nobody demos below the gate, whatever the commercial pressure.Work with our analytics engineers on the semantic layer and with our solution engineers on getting agents into the product behind a human gate.What we are looking forReal experience shipping LLM-based features to production, not notebooks. RAG, structured output, tool use, guardrails.Strong Python, strong SQL, and comfort on Databricks. Experience with Genie, Mosaic AI, or Unity Catalog is a strong plus. The Databricks Generative AI Engineer Associate certification is nice to have.You have built an evaluation set and measured a pass rate, and you can talk honestly about what it did not catch.Experience with agent frameworks (Claude Agent SDK, LangGraph, Semantic Kernel, or similar) and with designing the boundary an agent operates inside.Prompt engineering at expert depth. In most roles it is a starting competence. Here it is the craft.You understand that a semantic layer or metric definition is the source of truth, and you have worked with one.A background in supply chain, logistics, or operations data is an advantage. Knowing what OTIF or a dwell time actually means makes your evaluation questions better.A degree in Computer Science, Data Science, or a related field, or equivalent practical experience.Not sure you tick every box?Apply anyway. Nobody we have hired matched the list line for line. If you have put an LLM feature in front of real users and measured whether it was right, we want to talk to you. Tell us about the time your eval harness caught something, or the time it should have. That tells us more than a polished CV. Equal opportunitySwiftB2B hires on skill and curiosity. We do not discriminate on the basis of race, religion, nationality, gender, sexual orientation, age, disability, or any other protected characteristic. If you need an adjustment at any stage of the process, tell us and we will make it happen.