Transformations

Outputs matter. Judgment explains how they became possible.

These are not project summaries. Each story begins with the system I entered, what I observed before acting, the diagnosis beneath the visible problem, the trade-offs, and the capability I left behind.

3 flagship case studies
01Yarsa Tech · 2022–2023

Verified professional record

From fragmented teams to a functioning product organization

Hardware execution had stalled. The deeper constraint was not engineering capability; it was the system around it.

2 launches / 6 months15% lower product cost50,000-unit order
Read the transformation
02Healthcare AI · 2025–2026

Applied AI project

Designing AI around a real accessibility need

The harder question was not whether a model could classify sound, but how to make that information useful without adding distraction.

92% project benchmark19-paper reviewEnd-to-end prototype
Read the transformation
03Trustworthy AI · 2026

Working research

Studying when language models agree instead of telling the truth

A deterministic evaluation instrument for multi-turn sycophancy—without relying on another model as judge.

35 probes × 4 conditions10 repeat runs~1,400 API calls
Read the transformation

Evidence system

A result is only useful when its boundaries are visible.

01

Verified result

Supported by an approved record or public professional source.

02

Qualitative outcome

Supported by documented observation without an exact public metric.

03

Working research

An active study that is not presented as peer-reviewed or final.

04

Confidential detail withheld

Relevant evidence exists but ownership or disclosure limits publication.

The operating system

The stories change. The pattern is consistent.

Observe the system, diagnose the constraint, align the people, build evidence, execute the change, and leave durable capability.

Explore how I think