Agentic engineering
A practical workflow for using AI agents with architecture context, clear tasks, verification, and human review.
Architecture and product judgment remain human responsibilities in my workflow. Agents receive useful context, operate within clear constraints, and produce work that is tested and reviewed like any other engineering contribution.
1.Establish context
Start with the domain, system boundaries, repository conventions, and the intent behind the change. Make the constraints available alongside the code.
2.Scope the work
Define the expected behavior, acceptance criteria, and boundaries of the task so the output can be reviewed against a clear goal.
3.Verify the result
Review the diff, run the relevant tests and type checks, and exercise the changed behavior. Check the result against the original requirements.
4.Review and iterate
Engineers make the architecture and release decisions. Feed failures and review findings back into the next iteration before accepting the work.
Let’s talk about what you’re building.
I’m interested in fully remote architecture, principal engineering, and hands-on engineering leadership roles, especially with startups and mid-sized software companies.