The selected colleague will work at an MUFG office or client sites four days per week and work remotely one day. A member of our recruitment team will provide more details.
Prompt Engineer / LLM Specialist
OverviewJoin our innovative AI/DX Accelerator Integration team and play a key role in advancing artificial intelligence within the financial services sector. As a Prompt Engineer / LLM Specialist, you will design, optimize, and implement prompts, tools, and guardrails to support robust, compliant, and efficient AI solutions. You will contribute to the development of retrieval-augmented generation (RAG) systems and fine-tune large language models (LLMs), ensuring alignment with industry regulations and best practices for risk management, customer experience, and data security.
Key ResponsibilitiesPrompting & Evaluation- Develop and refine prompt architectures tailored to financial use cases, ensuring regulatory compliance and minimizing operational risk.
- Conduct A/B testing and automated evaluations using golden datasets representative of banking and financial scenarios.
- Collaborate with compliance and risk teams to validate prompt effectiveness and safety.
RAG & Safety- Design advanced retrieval systems (chunking, embeddings, rerankers) for secure and accurate information access in regulated environments.
- Implement policy-aware filters and automated redaction mechanisms for PII/PCI, adhering to FINRA, SEC, and GDPR standards.
- Work closely with cybersecurity and legal teams to ensure data protection and auditability.
Model Operations- Fine-tune LLMs for financial applications, optimizing for performance, compliance, and cost-effectiveness.
- Implement robust guardrails to mitigate hallucinations and ensure outputs are defensible and grounded in verified sources.
- Monitor and report on model performance, integrating feedback from end-users and compliance teams.
QualificationsRequired- 3+ years of hands-on experience with LLMs/NLP, including Python and modern machine learning tooling.
- Proven track record in deploying AI solutions within highly regulated environments.
- Strong understanding of data privacy, security, and ethical AI practices.
Preferred- In-depth familiarity with banking terminology, document types, and regulatory requirements (e.g., BSA/AML, PCI DSS).
- Experience building defensible AI systems with grounding, citations, and dynamic guardrails.
- Demonstrated ability to develop policy-aware retrieval and automated PII/PCI redaction patterns.
- Experience collaborating with compliance, risk, and audit teams in financial institutions.
Success Metrics- Consistently meet or exceed accuracy and safety targets as defined by financial industry benchmarks.
- Demonstrated reduction in hallucinations and false positives in model outputs.
- Positive user experience and product KPIs, such as improved task success rates and reduced time-to-answer.
- Acceptance of AI artifacts and guardrails by compliance, audit, and risk management teams.
- Full adherence to regulatory standards and industry best practices in all deliverables.
Education:- Bachelor's degree in Computer Science or a closely-related discipline, or an equivalent combination of formal education and experience
"Visa sponsorship/support is based on business needs. We do not anticipate providing visa sponsorship/support for this position."The typical base pay range for this role is as follows:
- New York / New Jersey: $ 100-171K
- Non-New York / New Jersey: $ 100-156K
depending on job-related knowledge, skills, experience and location. This role may also be eligible for certain discretionary performance-based bonus and/or incentive compensation. Additionally, our Total Rewards program provides colleagues with a competitive benefits package (in accordance with the eligibility requirements and respective terms of each) that includes comprehensive health and wellness benefits, retirement plans, educational assistance and training programs, income replacement for qualified employees with disabilities, paid maternity and parental bonding leave, and paid vacation, sick days, and holidays. For more information on our Total Rewards package, please click the link below.
Our hybrid work schedule is four days on-site and work remotely one day per week.
