chris fey contact

I build the systems that write the code.

Product engineer, at home in the backend. I design backends and the AI systems around them, and the tooling that lets a team trust what they produce.

Previously at Perspective Funnels, 2017–2026. First hire, then tech lead, responsible for the backend while engineering grew from 1 person to 25.

how I work

  • Product and engineering are one conversation.

    The best decisions I’ve been part of came from treating product and tech as two sides of the same coin. I’d rather shape the problem with the people who own it than pick up a finished ticket.

  • Guardrails, not gatekeeping.

    I work AI-first and rarely write code by hand anymore. My time goes into what makes the output trustworthy: evals, review gates, clear constraints, and skills the whole team can reuse.

  • Measure, then change course.

    At Perspective, the first approach to AI-generated pages needed around 40 model calls and still came out inconsistent. We replaced it with template selection and a content rewrite, backed by an eval suite. Being wrong early is cheap. Staying wrong is not.

  • Context is a budget.

    Model quality drops long before the context window is full, so I prune hard and keep each phase isolated. The same holds for people: gather the context first, then ask for someone’s time.

  • The parts that must not break should be boring.

    Thin adapters, no business logic in protocol layers, and separate infrastructure where the law asks for it. Clever belongs where it pays off.

selected work

At Perspective Funnels, 2017–2026

AI-generated funnels
Moved funnel creation from a JSON-driven Svelte app to complete AI-generated HTML pages, from hackathon prototype to production. Build time went from 9 minutes to 20 seconds; prompt caching cut input-token cost by about 90%.
A 4 TB MongoDB
Owned and operated the core production database for years, event sourcing included, through steady growth and the occasional live point-in-time restore.
MCP server for outside models
A thin protocol adapter with campaign creation as a server-side state machine, so external LLMs ask for what’s missing instead of juggling tools.
Shared AI workflows
A research, plan, implement flow, PR adjudication and cross-squad risk scans. Built for myself first, then handed to the team.

outside work

I’m drawn to the big questions: where technology takes us, how the universe began, what’s the best dad joke of all time. And to smaller ones, like the right order to watch every Star Wars and Marvel film, or the best strategy for planting strawberries.

let’s talk

I’m open to Staff and Principal engineering roles, and to freelance work in AI consulting, process optimization, and MongoDB performance and reliability.

hi@chrisfey.com

Near Frankfurt am Main. Happy to work remote.