Software Engineer · Full-Stack · End-to-End Ownership
Hand me a problem. I'll hand back a working product.
I work like an embedded product team of one: I take a problem from rough idea to a running product — design, build, deploy — with minimal hand-offs in between.
Operating Model
I tend to take a problem end-to-end — design, build, deploy, and the parts no one volunteers for: auth, billing, the deploy pipeline — so it lands as one decision instead of a chain of hand-offs.
I'm comfortable across the stack and across languages, so I can usually reach for the tool a problem wants rather than the one I happen to know. I lean on AI as a multiplier on delivery speed — it's a big part of why the two products below, each with auth, billing, real-time, AI extraction, and its own marketing site, shipped from one pair of hands. I've also led a small team, so I'm comfortable coordinating when a problem is bigger than one person.
Tier One / Two Products, One Builder
Tier One / Developer Tool, Shipped Solo
writequit.dev
Freelance developers juggle tasks, time, and invoices across three disconnected tools — and lose the thread.
- A full commercial SaaS — task manager, time tracker, invoice generator — in one terminal-native tool with vim keybindings and markdown tasks.
- Auth (WorkOS), invoice generation (jsPDF), real-time chat and shared client boards (Convex), password-gated permissioned sharing.
- The product design, the opinionated UX, and the marketing voice — not just the code.
- SvelteKit 5 · Convex · Tailwind 4 · deployed on Vercel.





A complete, billable product — auth to invoice to live client collaboration — shipped end-to-end by one builder.[Felipe: add metric — e.g., paying customers / time-to-first-invoice]
a tool, not a lifestyle · write. quit. ship. · for developers who'd rather ship than organize
Tier One / Consumer Product, Shipped Solo
pondera.casa
Choose the right house — without getting lost along the way.
Choosing a property means drowning in tabs and gut feeling. pondera turns a pasted listing URL into a structured decision — real scraping and LLM extraction, not a chat box — then curates, compares, and ranks.
Next.js 16 · React 19 · Convex · OpenRouter (gpt-4o-mini)
Paste
pondera.casa
Paste a link; AI extracts price, area, photos and a pros / cons / watch-outs card. No data entry. Score
pondera.casa
Score each property on a checklist; portfolio stats update as you go. Compare
pondera.casa
A side-by-side table with best value per criterion highlighted in gold. Decide
pondera.casa
AI ranks the options and explains each position in plain language. You decide.
Real AI work — scraping, extraction, curation, ranking — wrapped in a polished consumer product and its marketing site, shipped end-to-end.[Felipe: add metric — e.g., listings processed / decisions made]
Tier Two / Internal Product, Shipped Solo
Venture Lab — Niterra Group
A growing venture team ran on a CRM that couldn't carry portfolio reporting, expenses, or contracts. I built a custom automation layer on Affinity CRM, then scaled it into a full internal portal serving multiple business functions — a Portfolio Management module (automated quarterly report ingestion, real-time dashboards, custom reporting) and an Expense & Contract Management solution.
System map
Affinity CRM feeds an automation layer, which feeds an internal portal containing a Portfolio Management module and an Expense and Contract Management module.“One internal portal carried reporting, expenses, and contracts across multiple business functions — an automation layer built on Affinity CRM.”
Replaced manual quarterly report handling with automated ingestion and live dashboards.
[Felipe: add metric — hours saved per reporting cycle / # portfolio companies tracked]

Portfolio Management



Expense & Contract


Screens from a public demo seeded with fictional data — no real portfolio data shown.
Tier Two / Precision-Ag Platform · Web + Mobile
Flora — Santos Lab
Agronomists needed drone, satellite, and field data fused into something they could read on a map — and capture offline in the field. I built the React/TypeScript platform (Leaflet + Turf.js) and an offline-first React Native field app, and grew from developer into Tech Lead of five.




A web + mobile platform that put multi-source geospatial data in one place — and the case where I went from building to leading the build, developer to Tech Lead of 5.
Tier Two / Systems Behind the Screens
The systems users never see
HomeOptions' real engine was internal. For a call-center sales operation I owned a custom SvelteKit CRM wired into Twilio's call center, an AI-assisted call-handoff tool, and a Go/HTMX compliance & debt-settlement platform — the systems that don't screenshot well but run the business.
HomeOptions internal
A custom SvelteKit CRM integrates with the Twilio call center, an AI-assisted call-handoff tool, and a Go and HTMX compliance and debt-settlement platform.“Three internal systems for a call-center operation — CRM, AI call-handoff, and a compliance/debt-settlement platform — owned end-to-end.”
[Felipe: add metric — calls handled / agent handoff time / compliance cases processed]
Public-facing work



Back Matter / Built for the Love of It
When the work's done, I still build. A sampling of things made for the love of it — small games, starter templates, sprites and code, all mine.






Have a problem that needs an owner, not a committee?
Send me the messy version — I'll come back with where I'd start.
Felipe Afonso — Field Notes · A product team of one