A software architect in Singapore. I work in the AI platform team at GIC, helping scale and build business-focused AI agents.
Before that I led GIC's internal developer platform team and started an internal community for AI-driven development. Over more than 15 years I've designed systems and helped engineering teams change how they work, at ThoughtWorks, Standard Chartered, Citi and others. Outside work I build open-source tools for crews of AI coding agents, and write about what those agents need from the teams that use them.
Worked at and with
UNCHARTED
Notes from building with AI agents: how they remember, how their skills should be built and tested, and why teams need clearer process once agents join them.
If an agent retrieves the wrong document perfectly, does it have good memory? A vector database is an index, not memory. Agents need working, semantic and episodic memory kept apart, much as people do.
Read on LinkedIn → Article · Jan 2026Good engineers can compensate for loose structure; AI agents cannot. Why explicit workflows and an unambiguous definition of done matter more once agents join the team.
Read on LinkedIn → Article · Oct 2026Running a crew of coding agents turns a developer into a small engineering team. Seeing what that team is doing is the next problem.
Read on LinkedIn → Post · Jun 2026LLMs propose the changes, evals decide whether behaviour improved, humans review the diff. A skill becomes a small, testable package.
Read on LinkedIn →Shorter posts and earlier writing.
Graph-indexed code retrieval against plain grep on three real questions. Grep won on quality, speed and tokens; graphs may suit ranking and blast radius better.
Agent skills should be small, composable primitives with versions and dependencies. They need a package manager.
Three open-source tools for working with a crew of AI coding agents: a place to talk to the crew, a memory it can trust, and a harness that improves only with your approval.
Run a crew of AI coding agents without living in their terminals. One calm page on your own machine to talk to Firstmate, see what the crew is doing, answer its decisions and review real results.
npm install -g @chrisjowen/wheelhouse
Local-first team memory for software repositories, for people and agents alike.
npx @chrisjowen/docket setup
Make Claude Code better at working in your repository, one approved change at a time.
CLAUDE.md, skills, hooks and settings./plugin marketplace add chrisjowen/improve
Seventeen years of public code, from .NET and Scala through Elixir to today's agent tooling.
More than 15 years across consultancies, banks, media and insurtech: architecture, DevOps and coaching teams through change. At GIC that led from an internal developer platform and a community for AI-driven development to business-focused agents on the AI platform.
Roles since 2009. Full profile on LinkedIn
From my CV, current work, public repositories and certifications.
Education, certifications and the sectors I've worked in.
BSc Internet Computing, 2:1
University of Hull, UK
Consultancies – ThoughtWorks, McKinsey, Equal Experts
Banks – Standard Chartered, Citi, National Bank of Greece
Media and telco – Sky, Guardian, O2
Insurtech and startups – Uncharted and others
Advanced Kubernetes: 1 Core Concepts – LinkedIn, 2023
Build a Blockchain and a Cryptocurrency from Scratch – Udemy, 2021
Advanced Kubernetes Workshop – Learnk8s, 2019
Want to talk about AI platforms, agentic engineering or one of the projects?
The best way to reach me is on LinkedIn.
For the projects, open an issue on GitHub.