All work
Case study — Active

J — an AI engineering agent fleet

A personal multi-agent system that takes Jira tickets to reviewed pull requests

J — an AI engineering agent fleet
J — an AI engineering agent fleet — product preview
Role
Creator
Period
2025 — present
Stack
Claude Code Claude Agent SDK Bash GitHub Actions Jira

J is my personal AI engineering system: a fleet of Claude-powered agents that runs a real software delivery loop across the products I build.

How it works

A thin orchestrator routes each request to a specialized capability:

  • Estimate — pre-flight ticket sizing (XS–XL) with confidence scoring
  • Jira2PR — the one-shot flow: a Planner designs the approach, an Implementer writes the code, and a panel of independent Reviewer agents critiques it before the PR is finalized
  • Epic orchestration — wave-based implementation of whole epics on a feature branch
  • Review, merge, deploy — structured multi-reviewer code review, label-based bot merging, and production deploys with Jira transitions

Why I built it

Applying ten years of delivery-process design to agents instead of humans: enforced architecture principles, anti-pattern gates, audit trails, and a definition of failure that blocks bad code automatically. The interesting part isn’t the code generation — it’s the process engineering around it.