Systems Analyst · Backend & AI Systems
TypeScript / Node.js • Python • ERP Integrations • Regulated Systems
I build backend systems and integrations where traceability, consistency and operational continuity are real product requirements.
My work is centered on Node.js, TypeScript and Python, with experience in ERP integrations (SAP/TOTVS), APIs, data flows and regulated environments (BPF/GMP). I am expanding this foundation into AI systems: model integration, MCP, evaluation, guardrails, observability and secure operation.
I work as a Systems Analyst at Laboratório Cristália in the pharmaceutical sector. I contributed to the evolution of PhotoGIMP and today take part in organizing and reviewing the project.
| Project | What it demonstrates |
|---|---|
| Pimbas | Architectural evolution: domain modeling, contract preservation during migration, per-group data isolation. |
| Saturno | Product architecture: a modern Laravel/Inertia monolith with a modular boundary instead of premature distribution. |
| Maybe | Errors and absence as explicit states, adopted incrementally at the edges of PHP 7.4 systems. |
| Hush | RESP-compatible server: what concurrency, memory and TTL semantics have to make explicit. |
| MCP AnimagineXL | Applied AI engineering: exposing image generation over MCP and REST without hiding the GPU boundary. |
More dossiers and labs — including Resulta, Astrum and D4Sign Node — at gabrielalmir.com.br.
- Start from the domain and make trade-offs explicit.
- Prefer reproducible systems: tests, documented decisions, containers and clear operational limits.
- Treat AI as a system component that needs evaluation, access control, observability and human review — not as magic.
- Keep sensitive data, internal processes and credentials out of public repositories.
Q3 2026 — evaluation and guardrails for MCP servers; reporting pipelines over multiple databases.
- Reliable AI integrations with MCP, structured outputs and evaluation.
- Backend architecture for regulated and integration-heavy domains.
- Clear technical writing: architecture dossiers, limits and measurable outcomes.
TypeScript · Node.js · Python · NestJS · Fastify · PostgreSQL · Redis · Docker · AWS · GitHub Actions
"Não leio livros sobre o mundo. Saio por aí e vejo por mim mesmo." — Ezreal