AI Testing

Job Description BCE Global Tech's Global Quality Engineering (GQE) function is building one of Canada's most ambitious AI quality programs — certifying every AI and agentic system deployed across Bell Canada before it reaches production. As a QA AI Specialist, you sit at the intersection of artificial intelligence, software engineering, and quality assurance: a hybrid role that does not yet have a textbook, because the discipline is being written in real time. You will do two things simultaneously. First, you will bring AI into GQE's existing testing practice — embedding AI-powered capabilities into the test automation tooling, pipelines, and frameworks that 250 QA engineers already use every day. Second, you will build and operate the evaluation frameworks that test the AI systems being created by other Bell engineering teams — agents, orchestration pipelines, RAG applications, Salesforce AgentForce workflows, and ServiceNow Now Assist integrations. Requirements Key Responsibilities: 1. AI-Enhanced QA Tooling Modernize GQE’s QA stack by embedding AI to improve speed, coverage, and intelligence: Integrate AI-driven test generation into Selenium, Playwright, and Postman frameworks Use predictive models to prioritize tests based on code changes and defect history Enable self-healing automation for UI/API changes Automate defect triage and root-cause analysis using failure clustering Support natural-language test authoring (English/French) for non-technical QA Continuously pilot emerging AI testing tools via a technology radar 2. AI Evaluation & Quality Pipelines Build scalable evaluation systems tailored for AI behavior, not rule-based logic: Implement LLM-as-Judge pipelines on Vertex AI (Gemini) across key quality dimensions. Generate large, diverse, and adversarial test corpora from seed intents Evaluate RAG systems using metrics like faithfulness, relevance, and recall (RAGAS) Validate multi-step agent workflows, tool usage, and escalation behavior Embed AI evaluations into CI/CD as mandatory release gates. 3. AI Safety & Adversarial Testing Operate a dedicated AI red-teaming capability to uncover AI-specific risks: Execute prompt injection and poisoned-context attacks on RAG systems. Run automated jailbreak and constraint-bypass probes (e.g., Garak) Systematically test hallucination, numerical accuracy, and domain knowledge Assess toxicity, bias, and fairness across English and French interactions Stress-test agentic systems for runaway actions and scope violations 4. Continuous Quality Evolution Ensure the quality framework evolves as models and systems change: Monitor production AI outputs for quality drift and trigger re-certification Feed real production failures back into the test corpus Track model/version changes and generate quality delta reports. Maintain a living benchmark of Bell-specific AI quality standards Continuously adopt new evaluation research and industry best practices Partner early with AI/ML teams to embed quality by design 5. AI Quality Certification Operations Lead technical execution of the AIQC program: Own Tier 2 & 3 certification testing from corpus design to red-teaming Calibrate LLM-as-Judge rubrics using human-labeled golden datasets Produce clear AI Quality Certificates with scores, risks, and conditions Advise teams on AI testability, prompts, and evaluation instrumentation Contribute to AIQC playbooks, documentation, and knowledge sharing ​ Required ▸ 5+ years of software quality engineering experience, with at least 2 years working directly with AI/ML systems, LLMs, or AI-powered applications ▸ Hands-on experience building or evaluating LLM-based applications — including prompt engineering, RAG pipelines, or agentic workflows ▸ Proficiency in Python: test framework development, API integration, data processing, and evaluation scripting ▸ Experience with modern test automation frameworks (Playwright, Selenium, Pytest, RestAssured, Postman/Newman) and CI/CD platforms (GitHub Actions, Google Cloud Build, Jenkins) ▸ Working knowledge of at least one major AI/ML platform — Google Vertex AI, Azure OpenAI, or AWS Bedrock — with hands-on API usage ▸ Strong conceptual understanding of how LLMs work: tokenization, temperature and sampling, context windows, grounding, hallucination mechanics, and fine-tuning ▸ Demonstrated ability to design test strategies for non-deterministic systems — moving beyond assertion-based testing to probabilistic, rubric-based evaluation ​ Benefits What We Offer: Competitive salaries and comprehensive health benefits Flexible work hours and remote work options Professional development and training opportunities A supportive and inclusive work environment Access to cutting-edge technology and tools.