Shaping a 3D-print shop product and building a job-application pipeline
Turned a simple stock-control idea into a full product for a 3D modeling and printing shop, and built an end-to-end pipeline to draft and submit job applications.
A busy day across many workspaces, but two projects carried the weight: a custom product for a 3D print shop, and the tooling behind my job search.
3d-print-saas-management-system
This started as "a simple stock-control system" and grew into a full product for a shop that does 3D modeling and printing for clients. Across four workspaces I worked the scope from both ends: product and technical.
I split the docs into two layers that map to how I want agents to build it. A PRD describes the product (in pt-BR, since the client reads it), and a SPEC holds the technical plan that agents implement from. The domain took shape: clients, sellers, modelers, budgets, products, filament stock with cadastrable storage locations, production management with status, payment flow, demand queues with priorities, and a visual dashboard. Auth will use Supabase.
One useful reframing: it is not a SaaS for me to resell, it is a closed product for this one shop. That changes the cost math. I ran the numbers for up to ten concurrent users on Supabase and Vercel to find where the free tiers stop covering it.
The stack settled on Next.js, Supabase, shadcn/ui, and Bun. I also pulled in a set of skills for the build: frontend-design, ui-ux-pro-max, React helpers, shadcn plus Tailwind, and GitHub CLI, then pruned the ones that overlapped or conflicted.
career
The theme here was removing the manual grind from applying. I built an end-to-end application pipeline: it cuts source docs by relevance, generates a tailored PDF in a loop, and drives the actual submission. I tested it against real listings and even filled a real application form through the browser extension end to end.
The open question is the right interface. A browser extension that knows my profile and auto-fills personal data, cover letter, and the PDF upload feels like the better long-term shape than a one-off script per site. I also triaged several listings, saved the good ones to opportunities/, and dropped the ones that did not fit (a PHP role, for one).
Some days the product and the tooling teach each other: scoping a client's system and scoping my own job search are the same exercise in deciding what is actually worth building.