AD

Uses

A workflow designed for velocity, clarity, and fewer excuses.

I keep the stack practical. The goal is reducing friction between decision, build, and release.

Build stack

Next.js (App Router), TypeScript, React, and Tailwind for product-facing interfaces.

tRPC and Prisma for end-to-end type safety; FastAPI and async SQLAlchemy for Python services.

BullMQ + Redis for asynchronous pipelines, queues, and background AI tasks.

AI & LLM stack

Anthropic Claude SDK with LangGraph for multi-agent orchestration.

RAG with Voyage AI embeddings and pgvector for semantic search.

Evaluation harnesses for faithfulness, safety, and red-teaming before anything ships.

Infrastructure

PostgreSQL and Redis for persistence, caching, and queues.

Docker, Nginx + PM2, and Vercel for deployment.

Clerk for authentication and Stripe for billing.

Operating principle

Tools matter less than reducing friction between idea, decision, and release.

I optimize for leverage, not novelty for its own sake.

Built for signal

Strong products earn attention twice: once visually, then again through execution.

Start a conversation

Aveiro, Portugal

© 2026 Alex De Souza. Fullstack & AI Engineer.