New opportunity

Business Engagement and AI Solutions Lead

Marsh · Singapore

About this role

Marsh Risk seeks a Business Engagement and AI Solutions Lead for its Asia Risk Consulting team in Singapore. The lead will design and deliver enterprise-grade AI solutions for insurance broker risk consulting, connecting business requirements with solution architecture and overseeing delivery of production implementations.

Responsibilities

Lead engagement with Risk Engineers and subject-matter experts, run requirements workshops, capture user stories, and translate priority use cases into technical requirements.
Design AI and retrieval-augmented generation (RAG) capabilities. Produce high-level architecture and component designs covering ingestion, vector indexing, retrieval, LLM orchestration, backend APIs, frontend, storage, and monitoring. Define AI/RAG approaches, prompt-engineering playbooks, and technology choices.
Ensure solutions meet enterprise security, compliance, privacy, governance, PII-handling, and responsible AI standards. Design UX flows and wireframes for onsite applications; collaborate with engineers, lead code and architecture reviews, mentor teams, and maintain technical oversight during delivery.

Required qualifications

At least 12 years of technology experience, including significant hands-on solution architecture, AI/ML engineering, and system delivery experience; at least 5 years designing, developing, and deploying AI/ML solutions in production enterprise environments.
Expert-level Python and production-grade coding skills, plus proficiency in another language such as TypeScript, Java, Go, or Scala. Background in AI/ML frameworks, generative AI, RAG, NLP, and document intelligence; experience with AWS, Azure, or GCP, Docker, Kubernetes, CI/CD, and enterprise data architecture.
Strong stakeholder communication and workshop leadership skills. A bachelor’s or master’s degree in Computer Science, Information Technology, Engineering, Data Science, Machine Learning, or a related field.

Preferred qualifications include a relevant master’s degree; experience in insurance, financial services, or other heavily regulated industries; speech-to-text, OCR, or image-captioning experience; AI/ML research, patents, or open-source contributions; thought leadership; advanced MLOps and AI governance, including model versioning and experiment tracking; and relevant cloud ML or TOGAF certifications.

The role is hybrid in Singapore, requiring at least three days per week in the office or onsite with clients, including a team anchor day where applicable. Marsh describes professional development opportunities and benefits and rewards.

For full details about this role, click “Apply on company site” to view the employer’s original posting.

Skills for this role

PythonTypeScriptJavaGoScalaSolution architectureAI/ML engineeringGenerative AIRetrieval-augmented generationNatural language processingDocument intelligencePrompt engineeringVector indexingLLM orchestrationBackend APIsAWSAzureGoogle Cloud PlatformDockerKubernetesCI/CDEnterprise data architectureStakeholder communicationRequirements gatheringUX wireframingCode reviewResponsible AIAI governanceMLOpsModel versioning

Checking your profile…

Preparation suggestions, not the employer’s actual interview questions. Always check the original posting for current requirements.

Understand the role

What matters most

Turn risk-consulting needs into deliverable use cases
Lead workshops with Risk Engineers and subject-matter experts, capture user stories, and translate priority use cases into technical requirements for production implementations.
Own end-to-end AI/RAG architecture
Design ingestion, vector indexing, retrieval, LLM orchestration, APIs, frontend, storage, and monitoring; define AI/RAG approaches, prompt-engineering playbooks, and technology choices aligned with Risk Consulting needs.
Build in enterprise controls
Ensure solutions meet security, compliance, privacy, and governance standards, including PII handling and responsible AI practices.
Guide implementation through delivery
Design UX flows and wireframes for onsite applications, collaborate with engineers, lead code and architecture reviews, mentor teams, and maintain technical oversight.

Likely challenges to think through

Balancing use-case value with technical feasibility
Preparation hypothesis: SME priorities may not map neatly to feasible AI implementations. Prepare to explain how you would use workshops and user stories to clarify the need, compare solution approaches, and agree on technical requirements without promising capabilities prematurely.
Balancing retrieval quality with data protection
Preparation hypothesis: RAG design may involve trade-offs between useful retrieval and privacy or PII controls. Prepare an architecture walkthrough that connects ingestion, indexing, retrieval, and monitoring decisions to security, governance, and responsible AI requirements.
Maintaining delivery quality across teams
Preparation hypothesis: Broad architectural ownership and hands-on oversight may compete for time. Prepare to describe how component designs, code and architecture reviews, and engineering collaboration can keep a production implementation aligned without making every decision a bottleneck.

Practice your answers

Technical practice

How would you turn a Risk Engineer’s use case into technical requirements?

What to coverSuggested preparation (inference): Explain how you would capture the user story, clarify the intended decision or workflow, and agree on acceptance criteria with the SME. Distinguish business priorities from architecture choices before proposing an implementation.

How would you structure an enterprise AI/RAG solution for risk consulting?

What to coverSuggested preparation (inference): Walk through the components named in the posting: ingestion, vector indexing, retrieval, LLM orchestration, APIs, frontend, storage, and monitoring. Explain the boundaries between components and where security and governance controls would apply.

How would you prepare source documents for retrieval?

What to coverSuggested preparation (inference): Discuss document extraction, quality checks, metadata, and updates to indexed content. The posting names document intelligence and lists OCR as preferred experience, so treat OCR as a possible approach rather than an assumed requirement.

How would you choose and troubleshoot a retrieval approach?

What to coverSuggested preparation (inference): Compare indexing and retrieval choices using relevance, access controls, freshness, and operational cost. Explain how you would diagnose poor results by examining source content, indexing, and retrieved passages before changing the LLM.

How would you define an LLM orchestration and prompt-engineering playbook?

What to coverSuggested preparation (inference): Cover prompt purpose, context selection, output constraints, and a way to review changes consistently. Explain trade-offs between flexible answers and predictable, reviewable outputs for a risk-consulting workflow.

How would you evaluate and monitor a production RAG capability?

What to coverSuggested preparation (inference): Propose measures for retrieval quality, answer usefulness, reliability, and operational performance, with review against representative use cases. Connect monitoring to a process for investigating failures and deciding whether to revise data, prompts, retrieval, or architecture.

How would you handle PII and responsible AI in the design?

What to coverSuggested preparation (inference): Describe controls across ingestion, storage, retrieval, access, and generated outputs, including how you would check that sensitive information is handled appropriately. Explain how you would document and review risks against the enterprise security, privacy, compliance, and governance standards named in the posting.

What would you look for in a production-grade Python AI service?

What to coverSuggested preparation (inference): Discuss API boundaries, testability, error handling, observability, and maintainable code in an AI/ML service. Explain how code and architecture reviews would catch integration or reliability issues before production delivery.

How would you deploy and maintain the solution in an enterprise cloud environment?

What to coverSuggested preparation (inference): Explain how you would choose among cloud services and use containers, orchestration, and CI/CD while respecting enterprise data architecture. Address deployment and rollback trade-offs; model versioning and experiment tracking are listed as preferred expertise, not mandatory tooling.

How would you design the UX for an onsite AI application?

What to coverSuggested preparation (inference): Describe a user flow and wireframe that make retrieved information and generated output usable in the Risk Engineer’s workflow. Explain how you would validate the design with users and coordinate frontend and backend implementation with engineers.

Behavioural practice

Tell me about a requirements workshop with differing stakeholder needs.

What to coverSuggested preparation (inference): Choose a real example and describe the situation, how you elicited and reconciled needs, and the resulting agreed requirements. Focus on your actions and the outcome rather than claiming a particular Marsh workshop process.

Tell me about prioritising use cases when delivery capacity was limited.

What to coverSuggested preparation (inference): Use a real situation to show how you weighed business value, feasibility, and risk with stakeholders. State what you decided, how you communicated the trade-off, and what resulted.

Tell me about resolving an architecture disagreement during delivery.

What to coverSuggested preparation (inference): Describe the competing views, the evidence or design criteria you used, and your role in reaching a decision. Close with the effect on implementation or team alignment.

Tell me about mentoring engineers through a code or architecture review.

What to coverSuggested preparation (inference): Give a real example of the review context, the feedback you provided, and how you supported the team in acting on it. Describe the result without presenting routine review participation as a fabricated achievement.

Tell me about raising a security, privacy, or governance concern.

What to coverSuggested preparation (inference): Explain the situation, how you identified and communicated the concern, and how you worked with others to address it. State the resulting decision or safeguard, while distinguishing your own actions from the organisation’s processes.

Sharpen your knowledge

Topics to prepare

Risk-consulting requirements and use-case selection
The role connects Risk Engineers’ and SMEs’ needs to priority AI use cases. Suggested exercise: outline how you would compare two hypothetical risk-consulting requests by user value, data availability, delivery complexity and risk. These are preparation scenarios, not stated Marsh projects.
End-to-end RAG architecture
The posting explicitly calls for ingestion, vector indexing, retrieval, LLM orchestration, APIs, frontend, storage and monitoring. Suggested exercise: draw a component diagram and trace a document from ingestion to a user-facing answer, explaining the choices and failure points at each stage.
Generative AI and document intelligence
Generative AI, NLP, document intelligence and prompt-engineering playbooks are stated areas of work. Suggested exercise: compare how a proposed workflow would handle document extraction, retrieval and answer generation, then identify where prompt instructions help and where they cannot fix poor source data.
Security, privacy and responsible AI
Solutions must meet enterprise security, compliance, privacy and governance standards, including PII handling. Suggested exercise: annotate a hypothetical RAG data flow with access controls, PII exposure points, retention decisions and human-review needs; do not assume a particular Marsh policy.
Production architecture and delivery trade-offs
The role covers enterprise data architecture, cloud platforms, Docker, Kubernetes, CI/CD and production oversight. Suggested exercise: explain the trade-offs in deploying the same proposed architecture on one of AWS, Azure or GCP, including integration, monitoring and rollback considerations.

Skills to revise

Facilitate a requirements workshop
The lead runs workshops with Risk Engineers and SMEs and captures user stories. Suggested exercise: role-play a 20-minute workshop for a hypothetical use case, then write three prioritized user stories with acceptance criteria and unresolved questions.
Build and review a Python RAG slice
Expert-level Python and production-grade coding are required. Suggested exercise: implement a small Python flow for document ingestion, vector indexing, retrieval and answer generation; review it for error handling, tests and handling of sensitive input. Use a second stated language, such as TypeScript or Java, for a small API or interface if useful.
Write a prompt-engineering playbook
The role defines AI/RAG approaches and prompt-engineering playbooks. Suggested exercise: draft prompts for a hypothetical document-based answer, include instructions for insufficient evidence, and test them against relevant and irrelevant retrieved passages. This is practice, not a stated employer test.
Sketch onsite application UX
The posting asks for UX flows and wireframes for onsite applications. Suggested exercise: sketch a user flow from question entry to a retrieved-document answer, including how the user sees source context, errors and a route to flag an unsuitable response.
Conduct an architecture and delivery review
The lead reviews code and architecture, mentors engineers and oversees delivery. Suggested exercise: review your proposed solution as though advising an engineering team; record the highest-priority design and code issues, a cloud/container CI/CD delivery sequence, and the decisions requiring SME input.

Your preparation checklist

Map your evidence to the role
Prepare a one-page matrix linking examples you can substantiate to the stated requirements: 12+ years in technology, 5+ years delivering production enterprise AI/ML, solution architecture, expert-level Python, a second language, and the listed cloud and delivery tools.
Rehearse a production AI project walkthrough
Select a real project you can discuss. Prepare a concise account of the business need, your decisions and hands-on contribution, deployment, monitoring, results, and lessons learned. The role calls for both architecture leadership and delivery oversight.
Practice a Risk Consulting requirements workshop
As an interview exercise, draft an agenda, discovery questions, sample user stories, and prioritisation criteria for a hypothetical use case with Risk Engineers and subject-matter experts. Show how you would turn their needs into technical requirements.
Sketch an enterprise RAG solution
Prepare a high-level diagram for a hypothetical risk-consulting document workflow covering ingestion, vector indexing, retrieval, LLM orchestration, APIs, frontend, storage, and monitoring. Annotate technology trade-offs and a short prompt-engineering playbook; these are preparation examples, not a stated Marsh project.
Review safeguards and user experience
Add a simple onsite-application wireframe and a review checklist for security, privacy, PII handling, compliance, governance, and responsible AI. Be ready to explain how these considerations would affect the design and engineering reviews.
Prepare delivery and leadership examples
Outline specific examples, if available, of production-grade Python, another language, cloud deployment, Docker or Kubernetes, CI/CD, code reviews, and mentoring. Distinguish what you personally implemented from what you oversaw.
Write questions for Marsh
Bring questions about the Risk Consulting team's priority use cases, how Risk Engineers participate in requirements workshops, applicable AI governance constraints, and how architecture and delivery responsibilities are shared. Confirm the Singapore hybrid arrangement if relevant.

Availability last checked: 1 Oct 2026. Status can change after a check; confirm on the employer’s website. Report an issue · Our checking policy

KEEP EXPLORING

Similar AI jobs

Related skills and specializations in Singapore. Matched to this role, not your profile.

Explore more jobs

Finding similar opportunities…