I start with the user’s task and the parts of the process that can fail.
My background in import/export documentation keeps the focus on missing
information, exceptions and traceable records.
I use AI across research, implementation and debugging, then evaluate
the result against concrete checks. The work here includes editable
briefs, a reproduced duplicate-record failure and a fix accepted into
another developer’s project.
Direction & judgment
Turn the request into a focused workflow: the first action,
required inputs, failure cases and a clear handoff.
Execution & delivery
Use AI to implement and revise it. Check the working interface
or reproduce the bug, keep evidence and provide instructions
for the next person.
More experiments
QuoteLens ↗
A browser experiment in source-linked supplier quote
comparison, pack costs and review state. Independent project;
sample data is synthetic.
PourLab ↗
A water-sort puzzle studio with custom levels, manual play and
checked solution replay. An interaction and rule-correctness
experiment.