Forward Deployed AI Engineer, Full Stack

Forward Deployed AI Engineer, Full Stack AMZ Prep + Portfolio Companies | Full-time or Contract The short version We are hiring an AI-native full stack engineer to embed inside our business, find the manual work that is quietly costing us money, and replace it with production AI agents that hold up under real volume. This is not an R&D role and it is not an internal IT role. You will sit with the Finance, A/R, CSM, Freight, Operations, Sales, and Marketing teams, understand how the work actually gets done, and then ship the system that does it. The same person will also be deployed across our group of companies, so the systems you build for one business become the starting point for the next. Who and what we are AMZ Prep is a technology-driven 3PL and e-commerce fulfillment company operating 20+ warehouses across the US and Canada, moving 200,000+ units per day into Amazon for more than 1,300 brands. We combine fulfillment, freight, and logistics expertise with software that most of our competitors are still trying to buy off the shelf. AMZ Prep sits alongside several other ventures, and every one of them has the same underlying problem, which is a large volume of repetitive, judgment-light work that people are currently doing by hand. Solving it once, properly, and then redeploying that solution across the group is the entire thesis of this role. Our values Win Relentlessly: Move fast, stay resilient, and raise the bar.Own the Outcome: Take responsibility for what you build and fix problems when they arise.Be Humbly Confident: Communicate directly, collaborate well, and leave ego at the door.Obsess Over Data: Use data to make decisions, measure results, and continuously improve. Why the work is hard The models are not the hard part anymore. Anyone can get an impressive demo out of a frontier model in an afternoon, and most companies have a folder full of exactly that. The hard part starts after the demo, when the system has to handle a carrier invoice that does not match the rate card, a customer email that contains three separate requests, or a freight quote that needs a human in the loop because the numbers do not reconcile. Logistics is an unusually unforgiving environment for this. The data is fragmented across a WMS, a TMS, an ERP, carrier portals, marketplace APIs, and a hundred spreadsheets that someone maintains by hand. The workflows were designed around the people doing them rather than around any clean logical model. Volume is high enough that a system which is right 90 percent of the time creates more work than it saves. That gap between what the model can appear to do and what can actually be trusted in production is where this role lives. What you will own You own the outcome from problem to production. You work with the department, frame the problem, choose the architecture, build the system, ship it, and stay accountable for what happens after it goes live. Discover. You embed inside a department, learn how it actually makes or loses money, and find the work where AI creates real value rather than a nice demo. This means sitting with the A/R team while they chase invoices, riding along on freight quoting, and watching how a CSM handles a difficult account.Decide. You turn ambiguity into a technical plan and a build-versus-buy call, and you defend both when someone senior disagrees with you.Build. You ship the whole system rather than a piece of it, including the integrations, the interface the ops team will actually use, the evals, and the guardrails.Own. You measure what happens in production, you find out where the system is quietly failing, and you keep improving it long after the merge. When the scope turns out to be wrong, you are the person who says so and re-cuts it. This is closer to founding something inside a business than it is to picking tickets off a board. Where you will be deployed You will be connecting tools that do not talk to each other, building dashboards and insights, and shipping the agents that do the work itself. A real sample of what is on the list today: Finance and accounting An A/R agent that owns collections, chasing aging invoices and escalating only what needs a personA bookkeeping agent that accepts vendor invoices, codes them into QuickBooks, and issues customer invoicesA/P reconciliation against carrier billing, where the error rates are highest Business intelligence Health dashboards showing profitability by customer and margin by accountJoining data across systems that were never designed to be joined Customer success Inbound fulfillment request handling and day to day operational questionsCustomer service ticket triage and resolutionBilling agents replacing front-line task force workAutomated weekly reporting for every customer Operations Warehousing agents for inbound receiving, outbound order flow, and exceptionsB2B and retail compliance agents Freight Quoting, rate comparison, tracking exceptions, and carrier communication Sales and sales engineering Quoting agents that turn an inbound opportunity into a priced proposalContract review and redlining Marketing and outbound Outbound prospecting agents, inbound lead qualification, and calling agents Across the group Redeploying what works at AMZ Prep into the other companies If a person in this business is doing it today, it is in scope. The platform you help buildEvery engagement should make the next one faster. We want the agent scaffolding, the evaluation harnesses, the integration layer, and the internal tooling to accumulate into something that behaves like a factory rather than a series of one-off builds. You will help decide how that works, because it is early enough that the person who joins now shapes it rather than inheriting it. You will also get real latitude on tooling. We expect you to be orchestrating AI aggressively rather than typing every line yourself, and we would rather you ship three systems with strong judgment applied to generated code than one system written entirely by hand. Tech stack We run full-stack applications across multiple technology stacks - Angular + Java/Spring Boot and React + TypeScript/Node.js, on PostgreSQL. You will work directly across our production Java/Spring Boot and Node.js/TypeScript codebases from Day 1. Our own platform (Navigate) is built on Java/Spring Boot, so Java is a strong value add here - not a hard requirement, but a big plus. On the AI side we work with LLM orchestration, system prompts, context engineering, structured outputs, RAG, tool calling, and evals. Day to day you should expect to be using Cursor, Claude Code, GitHub Copilot, and similar tooling as a matter of course, and we work in GitHub with JIRA and Linear for tracking. Who we are looking for We care more about slope than years. The pattern we are betting on is a strong engineer who has shipped real production software, moved hard into AI, and improves faster than the people around them. That said, there is a floor, and it is a real one: You have shipped production software that real users depended on and that you stayed accountable for after launch.You have built something real with LLMs and can explain what made it reliable rather than what made it impressive. Shipped, not experimented with.You have genuine depth as a developer. You can read, debug, secure, and improve AI-generated code, and you know why it is wrong when it is wrong. We are explicitly not looking for someone whose entire experience is prompting a tool until something runs.You are strong with APIs, integrations, and messy production data, which is most of what logistics is.You have strong production experience with Java/Spring Boot and Node.js/TypeScript, with deep expertise in at least one of the two stacks.You have strong full-stack experience with a modern frontend framework such as Angular or React, using TypeScript.You are comfortable moving between Java and Node.js environments, and you know how to use AI to dramatically accelerate how you understand, build, test, and ship software.You have worked with GitHub and issue trackers such as JIRA or Linear.You can walk into an unfamiliar domain and be useful within days.You can explain a technical decision to a warehouse manager as clearly as you can to another engineer.You already work with coding agents and have real opinions about where they help and where they do not. What success looks like In the first 90 days you have embedded with at least one department, shipped one agent into production that the team actually uses without being told to, and can point to the manual hours it removed. By six months you have a second department automated and the shared infrastructure underneath both is starting to pay for itself. Within a year, the systems you built have been redeployed into at least one portfolio company. The measure is not how much you built. It is how much work no longer needs to be done by a person. The environment We move fast and the standards are high. The work is unusually close to the customer and to the operation, which means you will spend real time in warehouses and on calls rather than only in a repo. If you want narrow scope, a predictable pace, or distance from the business, this is not the right seat. This is in-person 3x per week in our Mississauga office. How to apply The process Intro conversation with our recruiting team.Conversation with the hiring manager about how you work and the decisions you have made.Paid work sample. We give you a real problem from our business, scoped to roughly a day, and you build against it. Details below.Systems design discussion where we work through the architecture and tradeoffs together.Final conversation with the founder. Our interviews are meant to feel like the job. There are no trick questions and no hidden criteria, so come ready to talk about tradeoffs and show us how you actually use AI in your work.