quietmason
Selected work

Three systems, described honestly

Client names, screenshots and metrics stay private by agreement. What follows is the engineering: what the problem was, what we built, and what it runs on. Details and code on request.

Operations systemin production

Replacing spreadsheets with one API

A multi-branch studio ran appointments, stock and commissions in spreadsheets. We built a web admin and a customer-facing Mini Program on a single API: booking with conflict rules, stock movements tied to service records, commission calculation, and monthly exports for the accountant.

Read the build notes →
React · Django · MySQLstack
Admin + Mini Programsurfaces
Role-based access for manager and staff. Reports the accountant can reconcile line by line.
Publishing pipelinein production

Dataset in, finished pages out

Content published by hand every week needed the repetition automated, without losing editorial control. The pipeline reads the source data, generates pages from templates, runs pre-publish quality gates, releases on a schedule, and reports a failure instead of shipping something broken.

Read the build notes →
Python · cron · static outputstack
Editors keep the decisionsdivision of labor
Gates block a release when a required field, image or link is missing — the failure is visible, not silent.
Storefrontin production

A store that had to be cheap to run

An English-language storefront where the catalog and every product page are generated from a single data file, order capture is wired straight to email notifications, and hosting has no monthly platform fee. A redirect map keeps everything that previously ranked alive.

Read the build notes →
Static output · PHP for formsstack
Zero platform dependencyhosting posture
Structured data and Core Web Vitals checked page by page before launch, not after.
How we describe work

No dashboards pointing upward

The rule these notes follow: if a number is not verified and approved by the client, it does not appear. If the client cannot be named, the work is anonymized without inventing detail: sector, constraint and process instead of a logo.

You will not find invented percentages here, and no client is named without consent. Every claim on this site is either something you can verify from a public source, or something we will show you privately when it is your project.

What we show

Architecture, decisions, trade-offs

The reason a pipeline retries three times, why a page is server-rendered, what we did not build. That is the part a buyer can actually evaluate.

What we withhold

Client identity and client numbers

Because a supplier who publishes your conversion rate to win the next pitch will publish the next client's too.