ZürichEngineering practice

Most companies don't need another agent pitch. They need someone who understands the business first — and then ships.

Shipwright is an engineering practice for teams whose technology and AI efforts are stuck between the idea and production — or whose teams have the tools and aren't getting real leverage from them.

Book a 30-minute conversationHow the Delivery Review works

The Shipwright mark: a sailboat over a sea of ones and zeros

25 years leading engineeringFractional CTO of a 2M-user platformShipping production code dailyLive enterprise agentic AI, shipped under regulatory requirements

Where are you stuck?

Three doors. Pick the one that sounds like your week.

Your AI initiative is stuck between the demo and production.

I've been on the other side of that gap — recently, under the strictest constraints.

Engineering

Your team has Cursor, Claude, Copilot — and velocity hasn't changed.

Two weeks to find out why, and fix it.

Consulting

You know AI matters and nobody on your team can tell you where to start.

I'll be that person. Part-time, no hype.

Advisory

Approach

The order matters.

Understand the business

What you sell, where the margin is, who signs off. Before any architecture.

Find where it's actually stuck

Usually not where the slide says. Data access, approval boundaries, or a team that was never given the time.

Ship

The smallest version that runs in production and survives an audit. Then the next one.

Start here

The AI Delivery Review

1–3 weeksfixed scopefixed fee

Where your AI effort is stalled and why, a review of architecture and tooling, and a concrete path to production or to team leverage.

You get a written report, a working session with your leads, and a recommendation you can act on without hiring me again.

Ask about the Review

Case storyTier-1 global reinsurer · anonymised

Situation: A claims-operations group wanted agentic AI in a live workflow, under FINMA-grade human-in-the-loop requirements. Two earlier pilots had not left the sandbox; the compliance function had stopped taking meetings.
What I did: Defined the target architecture around the approval boundary: which decisions a human signs, what gets logged, how the model is evaluated before and after release. Then led delivery of two phases with the in-house team.
Result: Both phases in production ahead of schedule. Audit-ready decision logs. The compliance function now attends the planning sessions.

Founder

Jason Angelus

Jason spent 25 years leading engineering — at Cisco, with organisations of 250+ engineers, and as fractional CTO of a community platform serving 2 million users. Since 2025 he has shipped production AI systems hands-on, including live enterprise agentic-AI work under regulatory requirements.

Jason works in English and Italian, with working German, from Zürich.

More about Jason →

Thirty minutes. Bring the thing that's stuck.

No deck, no pitch. You describe the situation; I tell you what I'd look at first.

Book a 30-minute conversation