Marius Argatu

I’m a Senior SDET with 10 years of experience building test infrastructure for systems where reliability matters, including airline payments, multi-tenant healthcare SaaS, fintech, and financial market data. I design automated frameworks, runners, and model-based test systems, with a track record of building QA processes from scratch. For the past several years, I’ve focused on testing AI-powered features, including LLM evaluation harnesses, RAG retrieval testing, golden-dataset regression, and agentic-system testing.

In my current role, I embed with engineering and own the quality strategy within my feature squad—defining what to test, at what level, and why. I maintain and extend API test frameworks, apply shift-left by authoring test scenarios during LLD design, and use agentic workflows for day-to-day test maintenance. I also built an internal LLM-powered debugging agent that correlates failing test output with Datadog traces, the current stack trace, and recent Jira changes—reducing triage time from hours to minutes, and I own regression planning and release-gate decisions for the payments domain.

Earlier, I built LLM evaluation for a RAG assistant answering researcher questions against FDA/EMA guidance using Langfuse and Deepeval, separating retrieval quality from generation quality with RAGAS and running evals in CI. I also created model-based testing systems on a permissions matrix with xState (generating roughly 1,000 scenarios from a single state machine definition) and built a Playwright + TypeScript platform using POM with fixture-based tenant/role isolation. Across roles, I’ve delivered automation frameworks in Python and supported both backend and UI test coverage—always aiming for test systems developers actually want to use.