I build the tooling that makes hard problems legible.

Cloud support engineer at AWS, working on generative-AI services. I reproduce before I answer, and I build the diagnostic tools I wish I had.

Work

Elsewhere

Portfolio

Three things I built because the tool did not exist.

Point at anything on the left for the short version. Each project page states the problem, the design decisions and why, and the tradeoffs I am not happy about.

Based: Sydney, Australia
Now: Support engineer, AWS
Before: Architecture, then design

Architecture write-up

Multi-agent diagnosis pipeline

A script holds the control flow; the agents do the thinking. Redundant parallel investigation, then reconciliation, then a build-then-independently-verify gate before a word gets written.

Stack: Python, shell, agent runtime
Scale: ~65 tool modules, ~1,900 tests
Code: Not public — internal systems

Write-up and source

Offline hybrid retrieval

Semantic similarity and SQLite full-text search, fused by reciprocal rank so two incomparable score scales never have to be normalised. Local ONNX embeddings, no network calls at all.

Stack: Python, ONNX, SQLite FTS5
Size: 554 lines, 2 dependencies
Index: 49 cases, ~3,400 chunks

Write-up and source

Self-checking deck engine

Slides as a plain data array, about forty lines of JavaScript, and a probe that measures every slide for overflow and reports it to a headless browser. Because "looks fine on my laptop" is not a test.

Stack: HTML, CSS, vanilla JS
Engine: ~40 lines, no dependencies
Checks: Per-slide overflow, in CI

Curriculum vitae

Experience, in full

Experience, education, certifications and skills, on screen or as a PDF. It prints properly, which took more effort than it should have.

Formats: HTML and PDF
Certs: 5 AWS, Apr–Nov 2025
Now: GenAI Developer – Professional

About

Architecture, then design, then cloud

The long version of a non-linear path: an architecture degree in Seoul, ten years of graphic and UX design across three countries, Python learned to contribute to a product API, then a deliberate retrain into cloud.

Languages: Korean, English, Portuguese
Studying: Dip. Cloud Computing, Swinburne

External

GitHub

Source for the projects that can be public, including this site and the confidentiality gate that keeps customer work out of it.

Handle: @Hee1ko

External

The professional record, and the better place to reach me about roles.

Handle: in/hee1ko

Contact

Say hello

Happy to talk about agentic systems, support engineering, or changing career into cloud. I answer everything that is not a template.

Email: hee1ko@gmail.com