Marius Argatu

Quality engineering for payment services and for LLM based tooling inside the delivery workflow.

• ▹ Own the quality strategy for my feature squad. What to test, at which level, and why. Developers write their own unit, integration and system tests; I review them and close the coverage gaps. Own regression planning and release gate decisions for the payments domain.
• ▹ Built an internal LLM powered debugging agent that correlates failing test output with Datadog traces, the live stack trace, and recent Jira changes, then iterates until the test passes or a real defect surfaces. Triage time cut from hours to minutes.
• ▹ Run agentic workflows in daily engineering: MCP servers for knowledge base retrieval, slash commands for bug triage, and agents for test maintenance.
• ▹ Maintain and extend the API test framework with fixtures, helpers and reusable modules, and drive coverage across new payment endpoints and business flows. Authored test scenarios during LLD design as the specification for developer written tests.

Regulated domain. Framework ownership and platform wide test architecture, plus the evaluation system for the product's RAG assistant and its agentic paths.

• ▹ Built the LLM evaluation system for a RAG assistant answering researcher questions against FDA and EMA guidance and tenant submissions. Separated retrieval quality (hit rate, precision@k) from generation quality (faithfulness, answer relevancy) using RAGAS. Golden datasets versioned in Langfuse, evaluation runs gated in CI.
• ▹ Extended evaluation to the assistant's agentic paths, covering multi turn state, tool and function calling and planning loops, so tool selection and trajectory could be scored and not only the final answer.
• ▹ Built a model based testing system on the permissions matrix using xState. Roughly 1,000 test scenarios generated from a single state machine definition, tracking the product as permissions evolve with no manual rewrite.
• ▹ Built a Playwright and TypeScript test platform with page object model and fixture based tenant and role isolation, adopted across multiple functional teams.
• ▹ Built an AST based migration tool that translated a legacy Karate DSL suite into equivalent TypeScript and Playwright code, removing a manual rewrite of thousands of lines.
• ▹ Acted as framework SME: onboarding, test pattern reviews, and coordination on cross cutting changes including an auth system rewrite.

Python and Pytest API framework, native desktop automation via patched Chromium.

• ▹ Owned which layers we tested at, which tools we standardised on, and how automation was structured across the native and embedded web surfaces. Ran technical workshops on mutation testing and property based testing.
• ▹ Built a custom Pytest plugin that runs dependent API tests concurrently with ordering resolved by a DAG. Runtime reduced from 1 hour to roughly 5 minutes.
• ▹ Patched and rebuilt Qt WebEngine's Chromium fork to expose the remote debugging port, replacing Appium with Selenium via CDP. Per test effort dropped from 4 days to 1.
• ▹ Introduced contract testing between the client and its backing services, moving a class of integration tests to a cheaper and faster layer.

• ▹ Designed a cross platform automation framework in Python using the Screenplay pattern, covering an SPA with Playwright and a Windows desktop client via WinAppDriver.
• ▹ Delivered an end to end suite that reduced manual regression on the desktop side, which had little automation before.
• ▹ Maintained and optimised Jenkins pipelines for execution and reporting across environments, and coordinated a small QA team.

• ▹ Led design of a Pytest based automation framework for complex ecommerce and payment workflows. The standards defined then are still in use.
• ▹ Built a database diff tool that captured state changes triggered by UI actions, letting QA snapshot and restore environments locally and validate results at the database level.
• ▹ Reduced test execution time through parallel execution and script optimisation.

• ▹ Shipped RESTful APIs in Python and Flask, validated performance under load with k6, and maintained CI/CD on Docker and Kubernetes.

• ▹ Owned integration testing of REST services end to end, and organised daily QA across web, desktop and mobile.

• ▹ Authored and evolved test cases against changing requirements, and ran regression cycles.

• ▹ End to end feature testing for AAA game titles across platforms.