AI and backend engineer. I build the system around the model.

ApproveHR · the assistant's query engine

Ask it about someone's salary. It depends who's asking.

The model never decides who may see what. On every request it's handed only the fields the asking person is allowed to see, and protected ones arrive redacted.

Also in production: a public-holiday engine that works out Nigeria's holidays from their rules and checks the news for the Federal Government's Eid declaration.

In review: a fairness check for performance ratings that hides groups under five, so no one person's score can be worked out, and AI review notes written with names taken out before the prompt.

Kowane · text-to-SQL evals

A query can run, return a number, and still be wrong.

Kowane answers a microfinance bank's questions in SQL. The hard part is knowing when it's wrong, so I built evals that send real questions through the real agent and score what comes back.

// The join was very confident. Confidence isn't a test.

And it runs on purpose: repeat runs to catch flaky answers, a baseline that fails the run on any regression, and a workflow to trigger before a prompt change or a model upgrade.

EDMS · lead backend developer

Every action leaves a link in a chain you can check.

Document management for enterprise and government. Each audit entry carries the hash of the one before, so a quiet edit breaks the chain.

And the AI part: scanned documents go through OCR, then Gemini writes a summary, behind a provider interface that can be swapped.

Android · debugging

My app's alarms kept vanishing. The phone was deleting them.

No crash, no error. The answer was one line, under a log tag I didn't write.