Kyle Naluan

A trading rule that looks good on a chart is usually a statistical accident. I hold every candidate to a pre-registered standard of evidence before it trades.

I found and fixed a backtest bug that manufactured a false trading edge, then wrote the fix into the project's own rules so it can't recur.

Correcting log footage by hand is repetitive and easy to get subtly wrong. I made correction automatic and camera-aware, and made the creative step reusable instead of one-off adjustments.

I ported the color engine to C++ and validated it against the original implementation, holding the two to about 1e-4 numerical parity.

Grinding problems and reading about concepts rarely sticks. I applied spaced repetition, retrieval practice, and interleaving, and defined “learned” as a measurable state, not a guess.

I withhold verdicts by default with a hint ladder - a design I pulled from a research review of learning-science evidence, not from personal preference.