RPA-GenAI Trust and Assurance Consultant

RPA-GenAI Trust and Assurance Consultant

• BTech/BE or equivalent technical degree. • 3–9 years of experience in testing/quality engineering, with hands-on exposure to Generative AI or LLM testing initiatives. • Working knowledge of AI testing concepts, including test design, evaluation metrics, and result interpretation for LLM outputs. • Experience contributing to structured documentation such as test plans, test evidence, and defect/risk reporting. • Proven experience executing red teaming for LLM applications and translating findings into actionable controls and mitigations. • Practical experience implementing Responsible AI practices and assurance workflows across the AI lifecycle. • Strong understanding of LLM failure modes (hallucinations, toxicity, bias, prompt sensitivity) and methods to test and reduce them. • Experience building automated evaluation pipelines and regression suites for LLM systems using DeepEval or similar frameworks. • Consulting experience: stakeholder management, requirement discovery, and delivering clear outcomes under ambiguity. • Define and execute testing strategies for Generative AI/LLM systems, covering functional, non-functional, and safety-focused validation. • Design evaluation frameworks and test suites using tools such as DeepEval to measure quality, robustness, and consistency of model outputs. • Plan and conduct red teaming exercises to identify vulnerabilities (e.g., prompt injection, jailbreaks, harmful content, data leakage) and recommend mitigations. • Establish Responsible AI assurance checks aligned to fairness, transparency, privacy, and safety expectations. • Create clear test plans, evidence, dashboards, and reports that communicate risks, findings, and remediation priorities to stakeholders. • Collaborate with engineering and product teams to integrate AI testing into delivery pipelines and improve release readiness criteria. • Support incident triage and root-cause analysis for model behavior issues, regressions, and evaluation drift across versions.