notvibescoded

Work

Case studies in craft.

Every project follows the same shape: the problem, the stack, the hardest part — that's where judgment beats raw generation — and the outcome.

Personal project · Next.js

Txtur

Problem

Clinics juggle appointments, patient records, and billing across disconnected tools — spreadsheets, paper, and memory. Txtur consolidates the daily workflow of a small clinic into one workspace.

Stack & architecture

Next.js frontend, structured data model, appointment scheduling, billing views. Designed as a coherent product, not a collection of generated screens.

Hardest part

The data model. Appointment states, billing line items, and patient history all interact, and getting the state machine right — what can transition to what, and what happens on cancellation — took more thought than all the UI combined. This is the part a prompt can't do: deciding what the system means.

Outcome

A working clinic workspace: scheduling, records, and billing in one place. AdSense and Search Console wired up; the product is live at txtur.me.

Data engineering · Scrapy + Django + React

Rental inventory pipeline

Problem

A client needed live rental inventory from dozens of NYC management-company websites — each with different layouts, anti-bot measures, and data formats — normalized into one feed for their platform.

Stack & architecture

Per-site Scrapy spiders behind a shared framework, upserting into a single normalized table; Django REST APIs serving units and building aggregates; a React dashboard on top. Scheduled runs, monitored for silent breakage.

Hardest part

Change detection. Sites redesign without telling you, and a spider that silently returns zero results is worse than one that crashes. The pipeline was built with canary checks and per-spider health signals, because the hardest bug in scraping isn't parsing — it's knowing when parsing stopped working.

Outcome

Hundreds of units flowing daily from the first three management companies, with a documented playbook for adding new ones. The pipeline survives redesigns instead of dying from them.

Mobile app · React Native

Carousels Maker

Problem

Creating swipeable carousel posts for social media meant fiddling with design tools that weren't built for the format. Carousels Maker is a mobile editor purpose-built for carousel slides.

Stack & architecture

React Native with free-layer editing, AI-assisted slide drafting, and one-tap export of single slides or full carousels as images.

Hardest part

The export pipeline. Rendering layered, editable designs to pixel-perfect images on device — with correct fonts, spacing, and safe areas across screen sizes — is fiddly, unglamorous work. No AI shortcut; just careful layout math and testing on real devices.

Outcome

A shipped app with AI-assisted drafting, free editing, and image export. Real software, on a real phone.

Have something that needs building properly?

No prompts-to-production pipelines. Just engineering: architecture, tests, review, and code you can actually maintain.

Start a project