Engineering principles
How I work
Concrete principles I apply both when building my own SaaS (VirCat) and when developing internal enterprise management systems.
Product first
Understand the business problem before writing code.
Before picking tools or designing tables, I map how the business actually operates and where the bottlenecks are. Concrete example: the purchasing, purchase management, returned deliveries and cash flow modules rebuilt at Importadora Garibaldi based on how the company really works.
Evidence-based quality
Tests before implementation (TDD), small changes via pull requests, review of every change, and testing in a real staging environment before production.
Quality is never left to chance. In VirCat I maintain over 1,200 automated test files and a pipeline where every change enters through a reviewed pull request and is verified in a dedicated staging environment prior to production.
AI with engineering judgment
I use AI agents to move faster, under strict controls: mandatory tests, adversarial code review and browser verification. AI executes; the decisions are mine.
AI agents multiply delivery speed only when paired with clear architecture and strict guardrails. I define the design, require automated tests, run adversarial review on every diff and verify the final behavior in the browser.
Scale by design
Systems designed to grow without one customer impacting another, with explicit limits on every query.
Running a multi-tenant platform in production teaches you that per-tenant index isolation, cursor pagination and strict query bounds belong in the initial design so a growing merchant never slows down the rest.
From real business needs to production
Whether building a SaaS platform from the ground up or rebuilding core modules inside an importing company, my priority is shipping maintainable software that solves the actual operational problem without introducing hidden debt.
- Domain understanding: I talk to the people who run the process every day before modeling data or defining API contracts.
- Short, verifiable cycles: I break work down into focused pull requests, backed by automated tests and validated in a separate staging environment before production deployment.
- Predictable operations: I design queries with explicit bounds, tenant-isolated indexes and clear production observability.