← All work & experience

Jan 2026–present / Project

Making agent work accountable.

AI Project System

Python · AI agents · Orchestration · Testing · Technical leadership

The context

An agent can report success without producing useful work. AI Project System explores how to organize software delivery around explicit scope, observable actions, reviewable artifacts, and human authority.

My contribution

The delivery workflow

Developed a governed workflow for defining work, dispatching agents, checking results, and recording what happened before deciding the next step.

Boundaries and review

Built bounded execution and retry paths, with human approval for consequential decisions and safeguards when required evidence is missing.

Experiments in public

Documented cases where successful process exits or confident QA verdicts did not correspond to completed work, alongside the evidence from a completed delivery run.

What came out of it

A public system with inspectable delivery records and documented failure cases. The completed Epic is evidence of one execution route; it is not a claim that arbitrary development tasks can run reliably without oversight.

Explore further

Back to this CV entryWork with me