Gabriel Carvalho

Managing personal finances demands spreadsheets, manual categorization of every expense, and constant discipline — most people give up before getting a real picture of their own money.

An AI-first financial organizer: you import bank statements (CSV/OFX/XLSX) and credit card bills in PDF, and the AI reads, reconciles, and classifies every expense automatically. The system shows the full picture — monthly and yearly balance, net-worth evolution — automates recurring entries, and guides you to your goals with budgets and month-by-month pace tracking.

• ✦Automatic classification with Anthropic Claude (BYOK — user-owned, encrypted keys) and PDF bill extraction
• ✦RAG with local embeddings on pgvector and per-merchant categorization memory that saves tokens
• ✦Hexagonal architecture (DDD) with NestJS on Fastify and Drizzle ORM, ready for multiple AI providers
• ✦Installable multi-user PWA with roles, statement-month handling for card bills, and timezone-correct reports

• NestJS
• Fastify
• Drizzle ORM
• PostgreSQL
• pgvector
• Next.js
• TanStack Query
• Anthropic Claude

Hotel chains need to operate several properties with different teams, tracking reservations, supplies, and cash — without any employee seeing more than their role allows.

A multi-tenant PMS (Property Management System) for the end-to-end operation: property registration, reservations, inventory and supplies control, cash flow, and team management. Each organization manages multiple hotels with account selection and role-based access control, with permissions per account and per property.

• ✦Multi-tenant with two-level RBAC: account permissions and per-property access
• ✦Full operation: reservations, inventory/supplies, cash flow, and staff in a single panel
• ✦Secure authentication with httpOnly cookies, CSRF protection, and rotating refresh tokens
• ✦DDD/hexagonal backend with NestJS + Fastify + Drizzle ORM and a versioned API

• NestJS
• Fastify
• Drizzle ORM
• PostgreSQL
• Next.js
• TypeScript

Recruiters and non-technical professionals need to evaluate candidates and job posts full of jargon — without knowing what each technology does or which roles use it.

A platform that translates technical jargon: it explains what each technology does, what it is for, and which professionals use it. It includes a prompt helper for building search terms, screening questions, and messages, making technical screening accessible.

• ✦Technology explanations translated for non-technical audiences
• ✦Mapping from technologies to roles and professional profiles
• ✦Prompt helper for searches, screenings, and candidate conversations
• ✦Fast, responsive interface built with Next.js 16 and React 19

• Next.js
• React 19
• TypeScript
• Tailwind CSS