Ismail Isleem

• Develop AI-powered test data generation systems using LLMs (ChatGPT & Claude APIs) with a dual-agent architecture — one agent generates synthetic positive/negative test data, a second validates outputs and provides structured feedback — using prompt engineering for deterministic, high-quality results.
• Build internal developer tools and Python testing utilities with LangChain, Hugging Face, OpenAI API, and transformers, exposing reusable libraries and CLI utilities consumed by automation engineers across the organization.
• Design internal QA service APIs exposing test data generation and dataset retrieval endpoints, applying DTO/POJO patterns for clean contracts; consumed by web and mobile automation pipelines.
• AI/LLM Testing Specialist: engineer custom evaluation pipelines for LLM-based systems — prompt testing, hallucination detection, context relevance & faithfulness scoring, RAG evaluation, semantic similarity via embeddings, and LLM-as-a-judge.
• Architect open-source test automation frameworks for enterprise clients across web, mobile, API, and desktop — coding directly in Selenium, Appium, Playwright, Robot Framework, RestAssured, and AutoIt.
• Lead in-house automation tools and QA platforms, directing a cross-functional engineering team through architecture, sprint planning, code reviews, and release readiness.
• Conduct yearly code audits across the organization and at major client engagements, evaluating framework design and design-pattern adherence (Factory, POM, Singleton); mentor engineers to raise code quality.

• Established the Quality Engineering function from the ground up within the AI/NLP department — defining testing strategy, frameworks, and tooling for Arabic language models and conversational AI, an emerging space with limited off-the-shelf Arabic tooling.
• Designed and built a full-stack ML model evaluation platform end-to-end: C#/.NET backend (ASP.NET, Code-First with Entity Framework) with a service-oriented API architecture for fault isolation, plus a JavaScript/HTML/Bootstrap no-code frontend used by manual QA engineers.
• Architected the evaluation engine supporting Word2Vec embeddings, POS tagging, NER, text classification, categorization, and normalization; ran datasets in parallel on Jenkins with configurable thread limits, computing Accuracy, Precision, Recall, F1 (Macro/Micro/Binary), and Confusion Matrix.
• Engineered web scraping pipelines for automated data collection, filtering, and categorization — generating internal training datasets that reduced reliance on third-party data purchases.
• Developed a rule-based Arabic text normalization and categorization engine in C#, used as a preprocessing layer for NLP models and a standalone service for the data science team.
• Extended the platform to test fine-tuned chatbots and conversational agents, including a production chatbot deployed for a banking client.
• Code review gatekeeper for the shared automation repository — approving all merges to master, running cross-division training, and mentoring manual QA engineers to run model evaluations independently.

• Owned full QA coverage (manual + automation) for an online vehicle auction platform serving the US and Canadian markets, reporting to the QA Manager.
• Rebuilt the automation infrastructure across three layers: web (Selenium/Java), mobile (Appium — Android & iOS), and JavaScript web testing (Cypress), applying reusable design patterns.
• Led the migration from Cypress to Playwright with Python, modernizing the web automation stack and improving stability and execution speed.
• Designed and built an in-house testing lab (dedicated servers and physical mobile devices) replacing cloud platforms — delivering significant cost savings while improving reliability and control.

• Owned the full QA lifecycle (manual + automation) for a logistics platform powering warehouse operations, slot management, inventory, and inter-service REST integrations.
• Designed two complementary API automation frameworks: a lightweight Postman collection for fast developer feedback, and a full-scale Java framework (Maven, TestNG, RestAssured) covering functional, regression, and smoke suites.
• Architected a cross-platform mobile automation framework (Appium, Java, Maven, TestNG) enabling unified Android/iOS execution.
• Migrated the suite to a BDD model with Cucumber, improving readability and validating integration flows between microservices for data consistency and contract integrity.

• Owned end-to-end QA for a high-traffic online classifieds platform — web functionality, user flows, and release reporting.
• Designed and implemented the company's first full web automation framework using Selenium and Java, establishing reusable components and structured test suites.
• Built API test automation for the search layer (Apache Solr) using Postman and JavaScript, and delivered weekly QA reports via Jira dashboards.

• Tested a US-based e-commerce platform and custom ERP covering warehouse management, inventory, and third-party shipping integration (DHL).
• Introduced the company's first automation initiative — Selenium/Java scripts for regression-heavy flows — transitioning into a hybrid manual/automation role; authored test cases and ran full regression cycles per release using Mantis.

Local-first AI tool that generates schema-valid synthetic test data through a dual-agent architecture — one agent produces positive and negative datasets, a second validates the output and returns structured feedback.

A shared core library powering three automation frameworks — web (Playwright), mobile (Appium), and API (httpx/GraphQL) — with parallel execution, reusable components, and unified reporting.