I question familiar ways of working, then solve problems with data and AI.

AX was progressing in scattered tracks — product, systems, automation. To bind them into one strategy I designed a Product/People/Process framework and structured the purpose, leading indicators, expected value and open problems for each axis.

Framework

3P + Foundation

Diagnostic

Level 1–5

Current → target

Lv2 → Lv3

• Set company-wide direction with the 3P AX framework
• Proposed the AX PMO organisation design
• Designed a Level 1–5 maturity diagnostic and presented it to the board

The company had to decide on a 450-person rollout. I designed and ran a survey across 89 engineers to establish the evidence, split 150 engineers and 300 non-engineers into 4 segments to estimate productivity savings, and compared 3 plan scenarios to land on a recommendation.

ROI

319%

Annual net gain

₩1.705B

Payback

2.9 months

• Designed and analysed a 89-person engineering survey for the evidence base
• Estimated savings across a job-type × usage-level segment matrix
• Compared three plan scenarios and put forward a recommendation

Formatting reports ate time every single week. I encoded the design rules and the writing standard as code, so Korean executive documents come out on-spec — then shipped it to the team.

Layouts

12

Component helpers

11

Form

Reusable skill

• Design system as code — tokens, 3 masters, 11 helpers, 12 layouts
• Font embedding implemented by editing the OOXML package directly
• Writing rules standardised, QA procedure built into the package

We needed to agree on how the organisation actually runs, but no existing tool shows role hierarchy and work cadence at the same time. I built the SVG renderer myself and wrapped it in an interactive report.

External libraries

0

Coordinate space

5 lanes × 4 cadences

Charts

4 from one builder

• Zero charting libraries — rendered directly via createElementNS
• 5 role lanes × 4 cadence bands, with semantically typed edges
• Four charts reuse one builder, so edits to the source propagate

Around ₩3B of marketing spend was being allocated on instinct. I built regression models to predict campaign performance and put them into the investment decision, then segmented customers and designed repeat purchase around it.

Repeat purchase

+20%

CPA efficiency

+11%

Spend designed

₩500M+

• Built campaign performance models on multivariate and logistic regression
• Designed repeat-purchase journeys per RFM segment — conversion +20%
• Allocated ₩500M of spend via MMM — CPA efficiency +11%

Data ran from sales to settlement but sat in separate places, so every meeting began with someone assembling it by hand. I mapped how the work actually flowed first, then built dashboards on top — so the evidence was ready when the meeting started.

Clients

30+

Reporting lead time

2 hrs → 15 min

Decision speed

+30%

• Mapped company-wide process from staff interviews → 7 dashboards
• Reporting lead time 2 hrs → 15 min (−87%)
• Decision support 30% faster, 30 minutes cut from daily meetings