QA Engineer for AI Products/Solutions

Hi, Please review the below JD:QA Engineer for AI Products/Solutions Location - Bellevue WAJob Description• Design and execute test plans for AI/ML-driven features, including model outputs, prompts, and integrated application behavior• Build and maintain automated test suites covering functional, regression, integration, and API testing• Evaluate model outputs for accuracy, consistency, bias, hallucination, and edge-case failures• Develop evaluation frameworks and golden datasets/test cases to benchmark model performance over time• Test prompt engineering changes, model version upgrades, and fine-tuning outputs for regressions• Perform adversarial and red-team style testing to surface safety, security, and robustness issues• Validate data pipelines feeding into AI models (data quality, schema, drift detection)• Collaborate with data scientists/ML engineers to define acceptance criteria and quality metrics for models• Test latency, scalability, and reliability of AI services under load• Contribute to CI/CD pipelines, integrating automated and model-evaluation tests• Hands-on experience testing LLM-based products (chatbots, copilots, RAG systems, AI agents) — designing test cases for non-deterministic, generative outputs• Practical experience with AI/LLM evaluation frameworks (e.g., Ragas, Deep Eval, Lang Smith, Promptfoo, OpenAI Evals, Tru Lens) — building eval suites, scoring rubrics, and golden datasets• Working knowledge of eval metrics for generative AI: hallucination rate, faithfulness/groundedness, relevance, answer correctness, toxicity/bias scoring, BLEU/ROUGE/semantic similarity where applicable• Experience with prompt regression testing — validating prompt changes and model/version upgrades against baseline eval sets• Strong proficiency in Python for writing eval scripts, test harnesses, and data validation logic• Familiarity with SQL and data validation techniques The environment is primarily Microsoft Azure-based, with Snowflake also playing a key role. Their GenAI initiatives largely leverage OpenAI and Claude models.