Backend Engineering
FastAPI and Spring Boot services, PostgreSQL models, validation and reconciliation engines, and retry-safe data pipelines.
Vinicius
Santana
Software Engineer — Backend & Full Stack
Backend engineer building Python/FastAPI and Java/Spring services for municipal public-health systems.
- Claim rejections cut from 12%+ to below 1% across 8 billing cycles
- Workforce reconciliation cut from 240+ to about 4 person-hours per cycle
Systems Atlas
Explore the systems, engineering decisions and evidence behind my work.
Brazil, UTC-3 · Software, data & operations
Four systems. A conceptual map of the work, not a live network.
Explore the atlas
Resume & contact
A quick path to the practical details.
I’m open to remote Software Engineering, Data Engineering and platform-focused conversations.
About
I build backend and full-stack systems for data-heavy, operational work.
I’m Vinicius Santana, a Software Engineer in Brazil working across Python, Java, TypeScript and AWS. I build FastAPI and Spring Boot services on PostgreSQL for municipal public-health operations—work that cut monthly manual review by 95% (21,000 to about 1,000 records across 17 facilities) and reduced workforce reconciliation from 240+ person-hours to about four per cycle.
Full Stack
Multi-tenant React/TypeScript dashboards with TanStack Query over FastAPI APIs, tested end to end with Vitest and Playwright.
Experience
Software, data, infrastructure and operations in practice.
Software Engineering & Health Data Systems at Prefeitura de Presidente Epitácio, Oct 2023 — Present.
- Built Python/FastAPI validation services and maintained Java/Spring/PostgreSQL systems, modeling longitudinal patient identity across historical names, missing or changed CPF identifiers, and CNS changes for 21,000+ monthly records.
- Reduced healthcare claim rejection rates from 12%+ to below 1% across 8 consecutive billing cycles by engineering 11 auditable Python/SQL validation and reconciliation rules across legacy and government systems.
- Cut municipality-wide workforce reconciliation from 240+ person-hours to about 4 per cycle by building Python extraction, Parquet processing, provenance controls, and centralized APIs.
- Built performance instrumentation to detect N+1 patterns, repeated execution paths, and latency regressions, validating changes through baseline/diff comparisons, shadow runs, and night/chaos tests.
- Diagnosed Java/Spring performance degradation caused by an unset optional identifier that prolonged object lifetime and increased garbage-collection pressure; fixed identifier assignment and cleanup behavior to restore stable execution.
- Debugged an edge agent across installation, discovery, and execution using isolated VMs, MITM traffic inspection, and runtime diagnostics; traced repeat-install failures to stale Windows Registry and machine state, then updated uninstall cleanup and forced discovery during installation.