Available for work
Carl-Arvid Ewerbring
I build production software with autonomous AI agents — and my job is to remove myself as the bottleneck.
About
I've spent about ten years in the game industry — as a developer, a producer, and a CEO — most of it in Unity. I'm a curious generalist who codes and loves to learn, and I'm as comfortable in a customer interview as I am in a terminal.
I like working with AI, and I'm looking for work in forward-deployed engineering — staying close to real users and real problems — or in applied AI. I like solving problems around how to use agentic AI.
Selected work
Five weeks ralph-looping a document extractiona and review tool
The ralph-loop from the last project, rebuilt into an authoring→build→measure pipeline that plans, refactors and audits itself — and then ported to run on Codex and Qwen. 36 days, 409 harness commits, 663 loop slices.
- AI
- case study
- internal tooling
- ralph-loop
- developer experience
Three days of research before a line of code
I pointed ~94 subagents at a lease-analysis SaaS I didn't know how to build. 3 days, 85 files, ~536k words, 690 sourced citations, 21 caught hallucinations, zero lines of product code.
- AI
- case study
- architecture
- research
- multi-agent
Five weeks to production with ralph-loops
A five-week, AI-native build of a production B2B booking system — 1,545 commits, 79.3% by autonomous agents, 63→2,196 tests, live. A work sample.
- AI
- case study
- agentic coding
- ralph-loop
Availability
I'm looking for forward-deployed or applied AI engineering work.
Small teams, smart colleagues, real AI budgets, and preferably direct contact with customers. I want to build things people actually want — not what I assume they need.
caewerbring@gmail.com