A year ago Rubyness was a Ruby on Rails company that also did local AI. We were good at both and comfortable with both. Then the job quietly changed. Clients stopped asking for "a developer who knows AI" and started asking for someone who could sit inside their business, understand the actual process, and ship a working AI solution into it. The industry has a name for that person: a forward-deployed engineer. We didn't choose the title. We grew into it.
What a forward-deployed engineer actually does
A forward-deployed engineer (FDE) is not a consultant and not a classic backend developer. They work next to the client's team, often inside the client's data and tools, and they own the result end to end: picking the model, wiring it into the existing stack, measuring whether it works, and fixing it when the real data turns out to be messier than the demo. For an AI implementation partner, this is the whole job. Slides don't ship. FDEs do.
From Rails and local AI to FDE: why the transition was natural
Rails taught us to care about the product, not the framework. Local AI taught us how models behave when they leave the sandbox: latency, privacy, cost, hallucinations on real documents. Put those two together and you get most of an FDE already. What we were missing was the habit of being embedded: working in the client's workflow instead of handing deliverables over the fence. That part took a year to build, and it was not a comfortable year.
The year that felt like a boot camp
We are honest about this with clients, so we'll be honest here. Twelve months ago, AI was a way to get your evenings back: let the model write the boilerplate and go home to your family. Then speed stopped being a feature and became a requirement. Everyone could ship fast with AI, so fast was no longer the differentiator. The differentiator became shipping fast and right, inside a real business, with real data, and being accountable for it.
So we went back to not sleeping. Work-life balance got worse before it got better. Every developer on the team went through the same training: real client projects, real deadlines, real failures when a pipeline that worked on test data fell over on a production export. We didn't run a course. We ran a year of production work and treated every project as the exam.
Why every Rubyness developer now works this way
The outcome is the thing we're proudest of. Today there is no "AI team" at Rubyness that other developers hand work to. Every engineer here can take an AI implementation from a conversation with your operations lead to a monitored system in production: scoping the use case, choosing between a hosted model and a private one, building the Rails or Python integration, setting up evaluation, and staying on when it breaks. That's what a forward-deployed engineer is, and that's now the baseline for everyone we put on a project.
For a client this means you get a senior engineer who can speak to both your CTO and your head of support, who has done this before on live systems, and who won't disappear after the handover.
Still growing, on purpose
The hard year paid off. We're taking on more AI implementation projects in the UK and Europe than at any point in the company's history, the team is growing, and we're hiring engineers who want to work embedded with clients rather than behind a ticket queue. We're keeping the bar where the year put it: every new developer goes through the same production-first path before they lead a client engagement.
What this means if you're choosing an AI implementation partner
Ask any vendor one question: who exactly will sit with my team, and have they shipped an AI system into production that looked like mine? If the answer is a solutions architect and a deck, you're buying a plan. If the answer is a forward-deployed engineer with a track record, you're buying the result.
At Rubyness, the answer is the second one, for every engineer on the team. If you have a process you'd like to see working with AI inside it, tell us about it. We'll start with your data and your workflow, not with a demo.