Key skills
AICommunication
About this role
Role Overview
- Reviewing AI-generated financial content and conversations for accuracy, appropriateness, and risk
- Identifying unsafe or misleading financial guidance (e.g., overconfident claims, risk minimization, inappropriate advice)
- Evaluating how AI systems handle uncertainty, disclaimers, and scope of knowledge in consumer-facing contexts
- Assessing whether models appropriately challenge or push back on risky or incorrect user assumptions
- Identifying patterns of sycophancy, over-alignment, or inappropriate agreement
- Evaluating multi-turn conversations for drift, escalation, and policy breakdown over time
- Participating in or reviewing red teaming exercises, including adversarial probing of AI systems to surface failure modes
- Evaluating how models respond under pressure, ambiguity, and escalating user intent
- Supporting quality assurance (QA) of red teaming outputs to ensure consistency and rigor
- Documenting edge cases, failure modes, and emerging risk patterns
- Providing structured written feedback to internal teams
- Collaborating with interdisciplinary teams on AI safety, policy, and evaluation frameworks
- Maintaining strict confidentiality and quality standards
Requirements
- Professional experience in one or more of the following: Finance, investment analysis, or financial advising
- Banking, wealth management, or asset management
- Financial risk, compliance, or regulatory roles
- Corporate finance, accounting, or financial planning
- Strong understanding of: Financial risk, uncertainty, and decision-making
- Appropriate vs. inappropriate financial guidance in consumer contexts
- Ability to identify: Overconfidence, misleading claims, or missing risk disclosures
- Inappropriate agreement with risky or incorrect user assumptions
- Failures in escalation, boundary setting, or uncertainty communication
- Strong written communication skills and ability to clearly explain reasoning
- Experience with or interest in: Red teaming, adversarial testing, or safety evaluation of AI systems
- Evaluating how systems fail under realistic user behavior
- Comfort working with AI tools and conversational outputs
- Ability to work remotely using Slack and standard productivity tools
- Comfort with ambiguity, iteration, and feedback-driven workflows
- Willingness to sign NDAs and work with sensitive content
- Availability ~10 hours per week for 8 weeks (starting in mid-April), with occasional scheduled meetings