I'm 21, studying software engineering in Brazil, and I've been building a portfolio of AI-powered micro-SaaS products by myself. No funding, no ads. Just SEO.
Four months in, I have 8 live products, about 1,000 programmatic landing pages indexed, and $500 in monthly recurring revenue. Most of that comes from 2 products. The rest are basically experiments that haven't proven anything yet.
Here's how the whole thing works.
The idea: take what works in English, rebuild it in Portuguese
Most micro-SaaS ideas I build already exist somewhere in the anglophone market. I find ones doing well on ProductHunt or IndieHackers, then check whether anyone has built the Brazilian version. Usually nobody has.
Brazil has 210 million people and the 5th largest internet population, but most global SaaS tools have zero Portuguese content. Not translated content, zero. So if I can build a decent version and write proper PT-BR landing pages before anyone else does, I own those keywords for months.
I won't pretend this is some brilliant insight. It's arbitrage. But it works, and it compounds.
How I actually build these things
Everything lives in one pnpm monorepo. Next.js with App Router, NestJS on AWS Lambda via SST, Supabase for the database, Stripe for payments. When I say "launch a new product," I mean I add a new package to the monorepo. Auth, billing, UI components are all shared. The actual new code per product is mostly business logic and landing page templates.
This setup is the reason I can ship a product in 1-2 weeks. Without it, I'd still be on product number 2.
The programmatic SEO part
This is the engine that makes the portfolio approach viable.
Take one of my products: an AI tool for real estate photo enhancement. Instead of writing one landing page and hoping it ranks, I pulled every long-tail keyword I could find from Google Keyword Planner in PT-BR. Things like "AI photo editing for [room type]" or "virtual staging [city name]." Then I classified them by intent, built page templates, and generated 200+ pages, each targeting a specific query.
I want to be honest about something here: this only works if each page is genuinely useful for the person who lands on it. I've seen programmatic SEO done badly, thin pages with swapped keywords and nothing else. Google catches that. My pages have unique copy, real examples, and the actual tool built in. It's more work per template, but the pages actually rank.
Across all 8 products, I have roughly 1,000 pages competing for long-tail queries where most competitors haven't written anything in Portuguese.
The products
PhotoGen is an AI photographer. You train a model with your selfies and it generates studio-quality photos. 145+ thematic packs, 7,000+ users so far. This one has the most traction.
AI ENEM generates practice questions for Brazil's national university entrance exam. Niche but the search volume is real during exam season.
Interior AI does AI interior design and real estate photo enhancement. The programmatic SEO example I described above is from this one.
Papagaio AI does AI video generation with lip sync and voice cloning in Portuguese. Still early.
Influenciador IA generates AI influencer content and virtual personas for brands. Testing product-market fit.
Rastreia Receita is a revenue tracking dashboard for Brazilian digital product creators who sell on Hotmart, Kiwify, or Stripe. Probably the one with the clearest B2B upgrade path.
Niverfy automates birthday reminders for teams. Simple product, simple value prop.
Pergunta PIX is an AI assistant for questions about PIX, Brazil's instant payment system. This one I'm not sure will survive.
Honestly, I expect half of these to die. The portfolio approach only works if I'm willing to kill the ones that don't get traction and redirect that energy to the ones that do.
What I've figured out so far
Doing one acquisition channel well beats spreading across five. I do zero paid acquisition. Everything is organic SEO. Month 1 was brutal because nothing ranks immediately. But I'm starting to see compounding effects now, and I don't have to pay for every visitor.
The monorepo saves me from myself. If I had to set up auth and billing from scratch for each product, I would have burned out after product 3. Sharing infrastructure is what makes the "build many, keep few" approach possible.
I do keyword research before I write a single line of code. I've killed at least 3 ideas that seemed exciting but had literally zero search volume in PT-BR. Would've wasted weeks on each one.
Writing in Portuguese, not just translating. There's a difference. Most AI tools have either English-only content or awkward machine translations. Writing native content means I'm competing against almost nobody for keywords that actually have volume. This won't last forever, but right now it's a real advantage.
And the uncomfortable truth: $500 MRR across 8 products means most of them are earning close to nothing. I'm being upfront about that because I've seen too many "I built X products" posts that hide the numbers. The revenue is concentrated in 2, maybe 3 products. The rest are bets that haven't paid off.
What I'm doing next
Cutting the products that show zero organic traction after 60 days of indexed pages. Expanding keyword coverage on the 2-3 that are working. Adding B2B pricing tiers to the ones with agency potential (Rastreia Receita is the obvious candidate). And shipping more products, because the math works better with more attempts.
I'll post updates here with actual numbers. If you have questions about any of this, or if you've tried programmatic SEO for your own stuff, I'd like to hear how it went.
I'm Marcelo Assis. I build AI micro-SaaS for the Brazilian market. You can find me on X or at marceloassis.com.br.
