This is the role our delivery model is built around. You will be embedded inside a client team for the length of an engagement, usually one to two at a time, and you will own the system from the first workflow audit through to the day it runs in production without you.
The work is remote-first but not remote-only. Audits and key build phases are typically on site, which means regular travel to client locations during an active engagement. How much varies by project, but a few days a month on average is a realistic expectation.
The scope is unusually broad. In one week you might map how an operations team actually processes exceptions, argue a director out of a project that will not pay off, then spend two days on the retry semantics of a tool-calling loop. If you like the engineering but not the room, this is not the right seat.
What you will do
- Run workflow audits: sit with the people doing the work, map what actually happens versus what the process document claims, and identify which workflow is worth rebuilding first
- Design and ship agentic systems that integrate with the stack a client already runs, without demanding a migration
- Own the production path: orchestration, evaluation, cost guardrails, observability, and the handover that lets the client team run it themselves
- Work directly with client engineers and stakeholders, including the ones who are sceptical that any of this will work
- Bring what you learn back into how we run the next engagement
What we look for
- Strong Python, and the judgement to know when a problem does not need an agent at all
- Production experience with LLM applications: retrieval, tool use, structured output, and the failure modes each of them brings
- Comfort with cloud infrastructure on AWS, Azure, or GCP, and with the CI/CD that keeps a deployment honest
- The communication skills to explain a trade-off to a CFO and a retry policy to a staff engineer on the same afternoon
- A track record of shipping something that other people depended on, and staying with it after launch
Nice to have
- Experience in non-software industries such as manufacturing, distribution, healthcare, or professional services
- You have inherited someone else's stalled AI pilot and got it moving
- Evaluation and observability tooling for LLM systems
Your first six months
- Month 1: shadow a live engagement and run your first workflow audit with a senior engineer
- Month 3: own an engagement, including the client relationship and the architecture decisions
- Month 6: your system is in production, the client team is running it without you, and you are shaping how we scope the next one