Backend / API
core- Python
- FastAPI
- Pydantic
Gabriel Sampaio · sampaiodev
I take a process out of improvisation and get it running.
I'm Gabriel Sampaio de Souza, a Python developer focused on automation, applied AI and infrastructure. I build systems with a real flow: input, processing, queue, database, logs, handled errors and useful output.
Trajectory
I learned by building things that must run
My focus is building what works in production, not what only looks good in the README. It can be a FastAPI API, a transcription pipeline, an AI agent wired to tools, an n8n flow or an app packaged for a Linux VPS.
The point is always the same: input, processing, persistence, diagnostics and useful output. I learned in practice, with open-source projects and real automations, taking care of the part that keeps the system standing — queue, logs, healthcheck and handled errors.
In June 2026 I started a degree in Systems Analysis and Development at Faculdade São Francisco de Assis to strengthen the formal foundation of what I already practice.
How I work
Architecture before the feature
I act as architect and reviewer, orchestrating AI agents — every change goes through plan, approval and implementation.
- 01PLANready
I plan the flow
Data, failures and operations mapped before any pretty screen.
- 02APPROVEapproved
I approve the direction
I review the architecture and only release it when it makes sense. Less rework, fewer surprises.
- 03IMPLEMENTlive
I implement and ship
Code that operates: logs, healthcheck, handled errors and reproducible deploys.
Stack & focus
Tools I actually use
Every tool is proven in a real project — grouped as subsystems of a single system.
Automation / AI
core- n8n
- LLMs
- RAG
- Whisper
Infra / Deploy
infra- Docker
- VPS Linux
- Dokploy
- GitHub Actions
Data / Queues
infra- PostgreSQL
- Redis
Product / Interface
surface- Next.js
- Streamlit
Focus areas
- Automation of real processes
- AI agents wired to tools
- Internal operations APIs
- Audio/video transcription and document generation
- Self-hosted systems and simple observability
Services
Where I solve the problem
A technical, operational offer: pull a manual process out of improvisation, integrate systems and ship to production. Not an agency package.
- 01
APIs and operational backends
FastAPI, Pydantic validation, auth when needed, operations endpoints, logs, versioning and deploy.
- 02
Automation with Python and n8n
Flows to cut manual work, integrate tools, standardize replies, generate documents and trigger tasks.
- 03
AI agents wired to tools
LLMs with structured prompts, RAG, knowledge bases, output schemas, external tools and simple guardrails.
- 04
Audio/video transcription pipelines
Google Drive, Deepgram, Whisper/faster-whisper, queues, export, webhooks and operation on a VPS.
- 05
Self-hosted infrastructure
Docker, Docker Compose, Dokploy, Linux VPS, Redis, PostgreSQL, volumes, healthchecks, backups and reproducible deploys.
- 06
Dashboards and operational integrations
Streamlit, Google Sheets, PostgreSQL, external APIs, internal reports and screens to track operations.
- 07
Technical review and architecture with AI
Stack diagnosis, monorepo organization, pragmatic Clean Architecture, basic security, observability and an execution plan.
Education
in progressFormal foundation in progress
Studying Systems Analysis and Development at Faculdade São Francisco de Assis since June 2026.
The degree reinforces what I already practice in real projects — it doesn't replace hands-on work, it adds to it.
Principles
What I believe when I build
- A system has to run outside the README.
- An error must show up before the client complains.
- Logs are for diagnosing, not for leaking sensitive data.
- Good automation reduces manual work without hiding risk.
- Simple infra that survives a restart beats pretty, fragile architecture.
Next step