Safwan Saba

Safwan Saba

AI Test Lead & Quality Engineering Architect

(352) 727-9967
safwan@safwansaba.infoPortfolioLinkedInGitHub

Professional Summary

AI Test Lead and Quality Engineering Architect with 14+ years of experience building quality strategy, automation platforms, and release gates for enterprise systems in finance, healthcare, and compliance. Hands-on in testing AI/ML and LLM applications: designing evaluation datasets and scoring rubrics, measuring accuracy, hallucination, and retrieval quality, and wiring continuous evaluation into CI/CD. Led teams of nine engineers across three offshore teams, and set the standards for reliable, trustworthy delivery. Creator of Sentinel, an AI-assisted self-healing QA platform, and an LLM-powered resume assistant validated with a structured evaluation suite.

Notable recognition: Discovered and responsibly disclosed a critical iOS security vulnerability (iOS 5.0.1) affecting 8M+ users worldwide.

Core Competencies

  • AI/ML Testing & Evaluation:AI test strategy, LLM evaluation, RAG evaluation, retrieval accuracy, hallucination and groundedness detection, prompt regression testing, evaluation datasets, scoring rubrics, accuracy, precision/recall/F1, robustness and adversarial testing, bias and fairness checks, drift monitoring, guardrail validation, human-in-the-loop review, agent workflow testing, Gemini API, OpenAI API
  • Quality Leadership:Test strategy and ownership, quality gates, acceptance criteria for AI features, release-readiness reporting, risk-based testing, team leadership and mentoring, cross-functional collaboration with data science, engineering, and product
  • Automation & Frameworks:Playwright (TypeScript, Java), Selenium WebDriver, Cypress, Appium, REST Assured, Cucumber/Gherkin, SpecFlow, TestNG, JUnit, Page Object Model, fixtures, network mocking, parallel execution, flaky-test remediation
  • Languages:Python, Java, TypeScript, C#
  • CI/CD & MLOps-Adjacent:Azure DevOps, GitHub Actions, Jenkins, Docker, Git, Maven, evaluation runs as pipeline quality gates
  • Distributed Systems Testing:Microservices, contract testing (Pact, consumer-driven contracts), gRPC, event-driven and asynchronous workflows, idempotency, retries, partial-failure recovery
  • Data, Cloud & Observability:MySQL, PostgreSQL, MongoDB, JDBC, AWS (EC2, S3, Lambda), Azure, Splunk, Grafana, Allure, Jira Xray

Projects

Neural Profile – Interactive Career Visualization Platform

Live Demo
  • Built an interactive neural network visualization of 15 years of career experience, where each skill is a node and connections represent relationships between technologies, projects, and expertise

Ask AI About Me – LLM Powered Interactive Resume

Live Demo
  • Built an interactive AI chatbot governed by system prompt engineering and Next.js context injection to accurately guide Gemini 2.5 Pro based on my professional background and skills

GlassBox – Next-Gen Autonomous QA Platform (Sentinel V2)

Live Demo
  • Designed GlassBox — transparent autonomous QA platform with 4 specialized agents in serial orchestration: Navigator (atomic browser action decomposition), Story Extractor (natural language → BDD scenarios), Persona Forge (semantic test data generation), Architect Weaver (strongly-typed POM + Playwright spec code generation)
  • Built Zero-Code AI pipeline using Gemini 2.5 Flash Vision: decomposes natural language directives into atomic Playwright browser actions with full reasoning logged per step for GlassBox transparency
  • Implemented Persona Forge: semantically infers form field intent from DOM context and autonomously generates realistic JSON test data profiles for Data-Driven Testing without any manual data setup
  • Designed Architect Weaver: AI code generator producing strongly-typed Page Object Models and Playwright spec files following SOLID principles — every generated file passes pre-flight build validation before committing
  • Architected ROI engine: 0.5 min saved per autonomous step + 15 min saved per healed selector, surfaced in live Mission Telemetry dashboard
  • Stack: Next.js 15, Playwright, Gemini 2.5 Flash Vision, Tailwind CSS, Genkit, TypeScript, Firebase/Firestore

Sentinel – Autonomous Self-Healing QA Platform

Live DemoArchitecture UIGitHub
  • Architected full-stack autonomous QA platform using Gemini 2.5 Pro and Genkit, reducing manual locator maintenance by 80% across production test suites
  • Implemented dual-agent Genkit healing pipeline: Analyst Agent reconstructs DOM state diff and generates the optimal CSS/XPath selector; XAI Auditor Agent produces a human-readable audit trail explaining every AI decision — enabling zero-trust transparency into autonomous fixes
  • Built Firebase-native RAG knowledge graph (Epistemic Journal) using gemini-embedding-001 (2048-dim, COSINE) — every healed selector is embedded and stored; future LLM calls retrieve top-10 semantically relevant memories before constructing any new selector
  • Engineered workspace-aware soft-boosting: same-site memories ranked higher, reducing cross-project selector hallucination
  • Engineered multimodal failure analysis pipeline ingesting DOM snapshots (1M token context), screenshots (base64 inline), Git diff (temporal regression), and stack traces — achieving 90%+ AI fix accuracy
  • Built Cortex Defect Intelligence Engine with 3-phase semantic target acquisition and headless ephemeral Docker probe runners that autonomously auto-close tickets in Firestore on verified fix
  • Built real-time mission telemetry with Firestore log streaming — stdout/stderr streamed line-by-line powering live dashboards with full status lifecycle
  • Implemented pre-flight build validation (mvn test -compile) ensuring AI-generated Java code never breaks the project
  • Engineered Epistemic Session Chaining (CICR) preserving auth tokens and session state across AI runs, enabling the platform to resume from last known state without re-authentication

Professional Experience

Senior QA Consultant | Asman Software Solutions

Jun 2025 – Present | Tampa, FL
  • Define AI testing strategy for client AI-enabled applications, covering functional, integration, and user-facing behavior, with entry and exit criteria for each release stage.
  • Design evaluation datasets and scoring rubrics for LLM and RAG features, measuring answer accuracy, groundedness, hallucination rate, and retrieval relevance, and track results release over release.
  • Run prompt and model-version regression testing, with evaluation runs gated in CI so quality regressions are caught before release.
  • Validate guardrails, adversarial and edge-case prompts, and human-in-the-loop review workflows for AI features.
  • Build Playwright and Python automation with reusable fixtures, API setup, network mocking, parallel execution, trace analysis, and deterministic test data.
  • Established consumer-driven contract testing with Pact to catch provider incompatibilities early and reduce late integration failures.
  • Use AI-assisted techniques for failure analysis and scenario generation, improving defect investigation speed and automation stability.

Lead Quality Engineer | PwC

May 2022 – May 2025 | Tampa, FL
  • Led and mentored nine engineers across three offshore teams, setting code-review standards, framework-design patterns, and testability practices that contributed to a 25% reduction in critical production defects.
  • Owned quality strategy for three enterprise applications: automation architecture, risk-based test strategy, release-readiness criteria, and quality reporting across UI, API, integration, and database layers.
  • Architected Playwright + TypeScript automation framework, increasing test coverage by 45% across web and API layers.
  • Designed Java BDD automation framework reducing regression cycles from 8 hours to under 3, saving ~20 engineering hours per sprint.
  • Drove CI/CD maturity in a SAFe environment by integrating 200+ automated API and database test cases into Azure DevOps pipelines, achieving 90% automation coverage.
  • Enhanced automation performance by 40% through MySQL and API testing strategy using JDBC and REST Assured.

Quality Engineer | Mednax

Aug 2021 – May 2022 | Sunrise, FL
  • Architected modular Page Object Model framework using Playwright with Java and TypeScript for critical healthcare applications
  • Boosted test coverage by 35% and eliminated 40% of manual regression effort through strategic automation
  • Implemented API testing strategy for healthcare data integration pipelines using REST Assured

Quality Analyst | PwC

Nov 2020 – Aug 2021 | Tampa, FL
  • Performed comprehensive end-to-end testing across front-end, back-end, API, and database layers
  • Developed C# + Selenium automation scripts using BDD (SpecFlow) within Azure DevOps

Quality Analyst | Fannie Mae

Jul 2020 – Nov 2020 | Tysons, VA
  • Developed automated regression suites using Java, Selenium, Cucumber, and Jenkins, reducing manual testing effort by 30%
  • Validated RESTful web services with Postman and REST Assured; performed mobile automation testing with Appium (iOS/Android)

Earlier Career Roles

2012 – 2020

Quality Engineering and Quality Assurance roles at Mobius Logic (Tysons, VA), Responsible Robotics (Tysons, VA), and MBC Group (Dubai, UAE). Responsibilities included building Hybrid Selenium frameworks, configuring Jenkins CI pipelines, and testing enterprise-scale media streaming platforms.

Education & Specialized Engineering Practice

Bachelor's Degree in Software Engineering and Automated Systems

State Engineering University of Armenia

2004 - 2008

LLM Agentic Systems — Applied Implementation

Designed and shipped multi-agent Genkit pipelines: dual-agent healing (Analyst + XAI Auditor), Firebase-native RAG with gemini-embedding-001 (2048-dim, COSINE vector search), multimodal Gemini 2.5 Pro ingestion (DOM + Vision + Git diff), and structured JSON output schemas — all in a production autonomous QA platform.

Deployed: Production (Sentinel Platform)