Sw engineer - ai first coding

The short version

We are building our next-generation business platform in-house, AI-first. You will be one of one or two developers doing it — working directly with the President, shipping software that the company runs on, every week. If you have taught yourself to build real systems with AI coding tools and want ownership rather than a ticket queue, this is an unusual seat.

About the role

This is not a traditional development role. You will work primarily through AI coding platforms (Claude Code or equivalent) to design, build, deploy and continuously improve the software that operates our business. The AI writes a great deal of the code; your job is to direct it, judge it, harden it, and own the result in production. The scope is the full lifecycle: cloud architecture on Microsoft Azure, module development, automation, rollout to the teams who use it, and the migration of work that lives today in an established legacy system and in spreadsheets. You will sit with the people whose jobs your software changes, and you will be accountable for whether it actually gets used.

What you will work on

• Build • Ship production modules across core business functions — inventory, purchasing, sales and operational reporting. • Own each module end to end: requirements, design, AI-assisted build, testing, deployment, iteration. • Review and rework AI-generated code to a production standard for correctness, security and maintainability. Knowing when not to trust the output is a core skill here.
• Data and integrations • Build and maintain data pipelines that move information between systems reliably and on schedule. • Work on financial workflows and controls where accuracy is absolute — figures must reconcile exactly, every time. • Build integrations for high-volume structured data exchange with external partners and internal systems.
• Platform and cloud • Design and maintain cloud architecture on Azure: environments, identity, access control, networking, deployment pipelines and monitoring. • Keep the platform secure and available — this is the system the business runs on during business hours.
• Automation • Design and deploy automation and assistant capabilities inside the platform that remove manual work and support decisions. • Measure whether they are accurate and useful, and improve them.
• Workplace technology — own it, then automate it • Own the company’s IT end to end: laptops and hardware, provisioning and deployment, accounts and access, software licensing, network and office infrastructure. • Be the place people come to when something does not work — and then remove the reason they had to come. • Simplify the stack: fewer tools, standard builds, one way of doing things instead of six. • Automate the repetitive half — provisioning, onboarding and offboarding, access requests, common fixes — using the same AI-first approach as everything else here.
• To be clear about what this is and is not: there is no IT department to inherit and we are not building one. The goal is a technology environment simple enough that supporting it is a small part of someone’s week, not a full-time job. You will feel the support load early; engineering it down is part of the work.
• Adoption and migration • Move processes off legacy tooling and manual workarounds, module by module, with data migration and parallel-run validation so nothing breaks. • Train and support users; stay close enough to the business to know what is actually needed.

The pace

This is the part most job descriptions get wrong, so it is worth being blunt: the delivery rate here is nothing like a conventional development team, and the expectation is set accordingly. • You ship in week one. Not a setup ticket — a real change, to a real environment. • You will ship something usable most weeks, and complete modules in weeks rather than quarters. • You will work across the whole stack in a single day: schema, backend, interface, deployment, and the conversation with the person who will use it. • You will move between business areas constantly. Breadth is the job; there is no single lane to settle into. • Some days the build stops because someone cannot work. Handling that well, and then automating it away, is part of the standard. The scope is open-ended by design. What is described above is what exists today — the platform keeps expanding into new parts of the business, and a good deal of what you build in your second year is not on any list that exists right now. If a fixed, well-defined remit is what you want, this will frustrate you.

What we hold constant is the standard, not the speed: it is correct, it is verified against real data, it is documented, and it works for the people using it. Fast and wrong is the one outcome we will not accept.

What we are looking for

Required • Demonstrated experience building real software with AI coding tools — you can show us what you have built and explain the decisions you made. • Degree in Computer Science, Software Engineering or equivalent practical capability. • Solid fundamentals: databases and SQL, APIs and integrations, version control, how a web application actually runs in production. • Cloud development experience; Azure, or the ability to become effective in Azure quickly. • The judgment to verify — you check results against real data instead of assuming the code is right because it ran. • Clear written communication with non-technical people. You will present to executives regularly.

Nice to have • One to two years of professional experience, co-op or internship experience. • Exposure to business systems, finance or operations data. • Experience building or deploying AI agents. • CI/CD and DevOps familiarity.

How we work — so you can judge the fit • Small. One or two developers and the President. No layers, no committee, no ticket queue. • Fast. Work ships to development continuously; production moves deliberately and on decision. • Evidence over opinion. Numbers reconcile or they do not. "It looks right" is not a standard. • Written down. Architecture and decisions get documented as they are made, not afterwards. • Honest. We say plainly what failed, what is unfinished, and what we do not know.