Sabino Software Ltda.

we build
systems
that have to work.

Sabino Software combines senior technical consulting with hands-on implementation. We work across data platforms, backend and distributed systems, applied AI, cloud infrastructure, and hardware-integrated products. Felipe Sabino stays accountable from framing through production and transfer.

Tell us what has to work Open the dossiers

Focused projects · embedded engineering · fractional technical leadership · longer engagements.

experience from working with teams at

Organizations shown reflect Felipe Sabino’s career history. Their names do not imply endorsement of Sabino Software.

  • Nubank
  • Conta Azul
  • Transfeera
  • Delivery Much
  • Alara
  • Flan Design
  • TeamCulture
  • Qexpert
  • Grupo Boticário
  • Fundação Estudar
  • Ambev
  • Accenture
  • NTT DATA
  • TOTVS
  • codeHB and Elastacloud
  • RHBrasil
  • Aragão Bus

production dossiers.

Seven engineering files from systems Felipe Sabino helped change in production. Public names stay visible; confidential work is described only through its engineering surface.

Transfeera dossier selected

CASE_001 / DATA PLATFORM

Transfeera

Principal Data Software Engineer
2023—2025

From a fragile EC2 pipeline to a declarative data platform.

Led a ground-up modernization that connected ingestion, transformation, orchestration, infrastructure, access, and local development into one maintainable platform.

  • Reframed mixed Airflow queries and Python jobs into a structured dbt-centered transformation layer.
  • Provisioned EKS and cloud infrastructure with Terraform, including a reproducible local workflow.
  • Coordinated a company-wide PostgreSQL WAL / CDC change and migrated streams through Datastream toward self-hosted PeerDB on EKS.
  • Improved BigQuery clustering and near-real-time delivery while mentoring engineers and guiding platform strategy.
CDCdbtBigQueryEKSTerraformPeerDB

systems shown by their engineering surface.

Some engagements can be described technically without exposing the organization or end deployment. Each extract below stays inside that boundary.

what we can
build with you.

You can bring Sabino Software into an architecture decision, an implementation, or a platform that needs to be recovered. The work stays connected: local development, production reliability, cost, documentation, and knowledge transfer all matter.

CAP_01

Data platforms & movement

Modernize pipelines, warehouses, lakehouses, CDC, metadata, governance, and analytics foundations without losing sight of reliability or cost.

  • dbt
  • Airflow
  • BigQuery
  • Spark
  • Databricks
  • PostgreSQL
  • Kafka
CAP_02

Backend & distributed systems

Design and build APIs, event-driven services, integration layers, high-concurrency workloads, and the operational tooling around them.

  • Python
  • Go
  • TypeScript
  • FastAPI
  • SQL
  • Kafka
  • REST
CAP_03

Applied AI & agentic products

Build RAG systems, tool-using agents, classifiers, evaluation loops, structured outputs, and human controls that make generative AI useful in real workflows.

  • LLMs
  • RAG
  • Agents
  • Retrieval
  • Evaluation
  • Human-in-loop
CAP_04

Cloud, platform & edge engineering

Make infrastructure reproducible, observable, and maintainable—from Kubernetes and Terraform to Linux, Android, hardware-integrated interfaces, and edge systems.

  • AWS
  • GCP
  • Azure
  • Kubernetes
  • Terraform
  • Docker
  • Linux
  1. STEP_01

    Frame

    Find the real constraint, risk, and definition of done.

  2. STEP_02

    Design

    Choose an architecture that fits the system and the team.

  3. STEP_03

    Build

    Implement the critical path and make feedback concrete.

  4. STEP_04

    Transfer

    Leave code, context, automation, and a team that can operate it.

open-source work and experiments.

These projects are how Felipe Sabino explores retro hardware, low-level constraints, data tooling, runtimes, and open-source collaboration. They are small enough for free experimentation and real enough to make every tradeoff visible.

SMBNeo running a platform game on Neo Geo NEO GEO · 320×224
PORT / CPlayable

SMBNeo

A pure-C Neo Geo port based on nathsou/smb, adapted for real hardware constraints with a browser-playable build. User-supplied game data stays local.

Play / projectSource
DoomGeo real-time raycaster running on Neo Geo M68K + Z80 · RESEARCH BUILD
RESEARCH FORK / CPrototype

DoomGeo

A real-time first-person raycasting experiment for Neo Geo, developed from NGRayEx. A study in rendering, memory, tooling, and honest hardware limits—not a claim of a full Doom port.

Research pageSource
OpenAPI FDW live request architecture
RUST / POSTGRESQLACTIVE

openapi_fdw

A native PostgreSQL foreign-data wrapper that turns OpenAPI-described JSON APIs into typed live tables, with JSONB evolution, transactional configuration, and opt-in writes.

Source
Open Platformer Runtime browser loader
GODOT 4 / C#EARLY PLAYABLE SLICE

open-platformer-runtime

A source-only runtime recreation for classic platformer projects, with a local ROM importer, browser loader, and deterministic verification tooling.

Try loaderSource
Doom64KB gameplay running on Neo Geo
NEO GEO / CPLAYABLE

Doom64KB

A Neo Geo edition of a real Doom port for 64 KB systems, including a sprite microframebuffer, multiple detail modes, streamed actors, and custom audio tooling.

PlaySource
BigQuery Cost Estimator extension preview
CHROME / EDGEPUBLISHED

BigQuery Cost Estimator

A local-first browser extension that adds cost estimates directly to the BigQuery console. Query text stays in the browser, with optional currency conversion.

Chrome Web StoreSource
dxt native transformation engine terminal
ZIG / DATA TOOLINGPRE-ALPHA

dxt

A Zig-first, dbt-project-compatible transformation engine focused on deterministic fixtures and artifact compatibility. A serious systems experiment, not a production recommendation.

Source

There is more in the workshop: data connectors, emulators, developer tools, infrastructure experiments, and contributions to larger open-source projects.

Browse GitHub

things worth sharing.

Workshops, releases, and engineering notes from the systems lab. The journal gives longer-lived context than a social feed and a home for what comes next.

Agent, MCP and RAG architecture from the 2025 workshop
AI Summit · Joinville

Do Prompt ao Protocolo: construindo agentes com MCP, RAG e IA aplicada

A hands-on workshop moving from prompts to tool-using agents, retrieval, MCP services, collaborative patterns, and an executable Docker-based reference architecture.

SlidesWorkshop code
Title slide for the 2024 resilient data extractors workshop
1st AI Summit · Joinville

Construindo extratores de dados resilientes* * mas não invencíveis! 😄

Why web extraction breaks, how LLMs can help interpret changing pages, and where deterministic validation still has to own the result.

SlidesWorkshop code
BigQuery Cost Estimator published extension
release note

BigQuery costs, visible before the query runs

The small browser tool is now published in the Chrome Web Store: multilingual, local-first, and explicit about what never leaves the browser.

InstallSource

More detailed build notes, workshop recaps, release decisions, and experiments will live here.

Open the journal
Felipe Sabino
Felipe SabinoFOUNDER
PRINCIPAL ENGINEER

i still design,
code, and ship.

I founded Sabino Software because I do my best work close to the system: understanding what is wrong, choosing an architecture, writing the code, and helping the team own it afterward.

Over 19+ years in professional software engineering and roughly a decade in production data platforms, I have worked inside startups, scale-ups, consultancies, and large enterprises, including Nubank, Conta Azul, Transfeera, Accenture, NTT DATA, and TOTVS.

My range now includes distributed data, backend services, cloud platforms, developer experience, applied AI, and hardware-integrated software. I still prefer the same working pattern: make the problem concrete, build the critical path, explain the decisions, and leave the system easier to operate.

  • principle_01

    make decisions executable

    The important choices become clearer when the design can actually run.

  • principle_02

    make the system understandable

    Documentation, observability, and a usable local workflow are production features.

  • principle_03

    leave ownership with the team

    Good delivery creates capability instead of dependence on the consultant.

  • principle_04

    understand before replacing

    Modernization starts by learning why the current system became difficult.

Ways to work together

Focused technical delivery · platform modernization · fractional technical leadership · embedded engineering · architecture and implementation reviews.

tell me what
has to work.

No public inbox and no automated sales sequence. Give me enough context to understand the system, the constraint, and what a useful outcome looks like.

or find me on LinkedIn ↗
human check initializes here

private by design · protected by Cloudflare Turnstile