FNP The practical side of my LinkedIn.
On LinkedIn you see where I worked. Here you see what I built on my own. Three products, from the first idea to real customers. Each case shows how I work as a PO and BA.
Language lessons on WhatsApp
The problem
Most people quit language apps after a few weeks. But everyone opens WhatsApp every day. So I put the teacher inside WhatsApp.
What I did
- Designed the method. One short lesson a day with vocabulary, grammar, a dialogue, audio and two tasks.
- Planned a 365 day curriculum per language and produced 1,825 lessons with AI. Then I reviewed them.
- Specified the full flow. Sign up, 7 day trial, daily lesson in the student's time zone, AI correction of text and voice, spaced review and payment.
- Built it on n8n and PostgreSQL with the WhatsApp API. No human in the daily operation.
- Added monitoring. Alerts for errors and AI usage limits, plus a daily summary on my phone.
Key decisions
- One time payment instead of a subscription. It is easier to sell in Brazil.
- Voice tasks with spoken feedback. It is the feature people remember.
- Free AI tiers with an alert at 80% of the limit. The running cost stays close to zero.
Where it is now
Live with paying students. Licensed as white label to a language school, with its own number and payment account. Next is a voice conversation product as an upsell.
Built with
- n8n
- PostgreSQL
- WhatsApp API
- LLM
- Whisper
- TTS
Escreva 3 frases em italiano usando “c'è” ou “ci sono” pra descrever o que tem dentro da sua bolsa agora.
Scheduling and CRM on WhatsApp
The problem
Small barbershops and clinics run their schedule on paper and on WhatsApp. No shows cost money. Owners have no clear view of revenue or commissions.
What I did
- Ran product discovery and wrote the product vision with four personas.
- Set three product principles. Progressive complexity. WhatsApp at the core. Phone for staff, desktop for owners.
- Split the roadmap into 8 sub projects. Each one went through spec, plan, build and tests.
- Designed for Brazil and Italy from day one. Two languages, two currencies, LGPD and GDPR.
- Built the app with AI pair programming on Next.js, PostgreSQL and the WhatsApp API.
Key decisions
- Team screens only appear when a second professional joins. A solo barber never sees them.
- No online payment for end clients. WhatsApp confirmation and reminders reduce no shows instead.
- One plan at R$ 49.90 per month with a 14 day free trial.
Where it is now
Built in 24 days and launched in Brazil with a free trial.
Built with
- Next.js
- TypeScript
- PostgreSQL
- Prisma
- WhatsApp API
- Docker
| Persona | Device | Needs |
|---|---|---|
| Owner | Desktop | Everyone's schedule, revenue and commissions |
| Professional | Phone | Today's schedule, time off and own clients |
| Front desk | Desktop | Book and reschedule, without finance |
| End client | Phone | Book in 4 taps and cancel from WhatsApp. No login. |
ERP for small businesses
The problem
Small businesses use separate tools for sales, invoices and finance. Data is typed twice and numbers do not match.
What I did
- Mapped the processes and merged 4 separate tools into one platform.
- CRM, Brazilian e-invoicing, payables, receivables and inventory in one place.
- Built with Claude Code from my own specs.
Where it is now
In testing with pilot companies.
Built with
- Claude Code
- PostgreSQL
- Docker
Also in progress
Summa
Personal finance app. Prototype.
Mercado
Grocery app. Scan a product, get a health score and track prices from your receipts. Prototype in Flutter.
Tudia Conversation
A daily voice chat with an AI character in the language you study. Design ready.
How I work
The same method in my own products and at my clients.
Discovery
I sit with the stakeholder, map the process and agree on what is in scope and what stays out.
Written spec
Every feature starts as a short spec with the decisions and the reasons. 18 specs across these projects.
Plan
The spec becomes a step by step plan that a developer or an AI can follow. 21 plans so far.
Build with AI
I direct the AI that writes the code and I review the result. Tests run on every change.
Measure
Alerts, dashboards and real user feedback decide the next step.
I used the same approach at SRS Distribution. A product I specified was built with 80% AI generated code, in 1 sprint instead of 3.
Hiring a PO or BA?
I live in Trento and hold Italian citizenship. Open to remote, hybrid or on site roles.
fnptecnologia@gmail.com