New opportunity

Information Technology - Senior Data Scientist (Advanced AI Track)

Singapore Airlines Ltd · Singapore

About this role

Join Singapore Airlines' in-house AI Center of Excellence to build and operate shared AI platforms, including LLM gateways and agent orchestration layers. You will deliver LLM and agent applications from prototype to production, ensuring measurable business impact. Key Responsibilities: Build and maintain shared AI infrastructure, including evaluation and monitoring layers. Develop agentic workflows with a focus on tool-call accuracy, policy compliance, and human-escalation effectiveness. Balance cost and performance by routing between models and managing GPU resources. Design knowledge retrieval systems with strict access control and data governance. Apply traditional machine learning for fraud detection, forecasting, and resource optimization. Collaborate with Data Engineering and Enterprise Architecture teams to lead end-to-end AI projects. Requirements: Degree in Computer Science, Machine Learning, or related field. Minimum 3 years of experience delivering ML or AI systems to production. Proficiency in Python and SQL. Expertise in traditional ML techniques (gradient-boosted trees, regression, clustering, etc.) and LLM applications using commercial or open-weight models (Claude, GPT, Gemini, AWS Bedrock). Strong understanding of structured evaluation, monitoring, and responsible AI practices including guardrails, groundedness checks, and data privacy (PDPA). Experience with secure cloud deployment (AWS, Azure, or GCP) and cost management. Strong project management, communication, and mentoring skills are essential.

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

Skills for this role

PythonSQLMachine LearningLLMAI AgentsPrompt EngineeringRetrieval-Augmented GenerationAWSAzureGCPGradient-boosted treesRegressionForecastingClusteringAnomaly detectionFeature engineeringModel evaluationResponsible AIData privacyCloud deploymentCost management

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

Production-Grade AI Infrastructure and Evaluation
The role requires building and maintaining shared AI platforms, specifically LLM gateways and agent orchestration layers. You must prioritize the development of robust evaluation frameworks that measure groundedness, latency, and reliability, as the team is responsible for moving prototypes to live services with measurable business impact.
Agentic Workflow Reliability and Compliance
A core priority is ensuring agentic workflows function safely and accurately. This involves rigorous testing of tool-call accuracy, enforcing policy compliance, and managing human-escalation effectiveness to prevent unauthorized actions within the enterprise environment.
Strategic Model Selection and Cost Optimization
You must balance performance with cost by intelligently routing between commercial models, open-weight models, and internal GPU resources. This requires a strong ability to determine when traditional machine learning techniques are more efficient than LLMs for specific tasks like forecasting or anomaly detection.
Cross-Functional Project Leadership and Governance
As a technical lead, you are expected to collaborate with Data Engineering and Enterprise Architecture teams to manage end-to-end projects. This includes designing knowledge retrieval systems that strictly adhere to data governance and privacy regulations like PDPA.

Likely challenges to think through

Navigating Ambiguous Business Requirements
The role requires translating vague business requests into technical solutions while managing stakeholder expectations. Prepare by practicing how you define success metrics for AI projects and how you communicate technical limitations to non-technical business units without using jargon.
Maintaining Security and Privacy in LLM Deployment
Implementing LLMs within a large organization like Singapore Airlines involves strict data access controls. Prepare by reviewing best practices for implementing guardrails, groundedness checks, and audit trails to ensure sensitive corporate data remains secure during retrieval-augmented generation processes.
Managing Model Drift and Lifecycle
The responsibility for model upgrades and retirements in a production environment is significant. Prepare by articulating your approach to monitoring model performance over time and establishing clear release gates and rollback criteria to maintain system stability.

Practice your answers

Technical practice

Designing LLM Gateway and Agent Orchestration

What to coverExplain how you would architect a multi-provider LLM gateway to manage routing, latency, and cost. Discuss the trade-offs between using commercial APIs versus open-weight models and how you would implement an orchestration layer to handle complex agentic workflows.

Structured Evaluation of LLM Systems

What to coverDescribe your approach to building test sets for evaluating LLM applications. Focus on how you measure groundedness, factual accuracy, and task-success rates, and how you define release gates to prevent model or data drift in production.

Agentic Tool-Call Accuracy and Compliance

What to coverHow do you ensure agents adhere to policy compliance and maintain high tool-call accuracy? Discuss strategies for preventing unauthorized actions and designing effective human-escalation triggers when an agent fails to complete a task.

Knowledge Retrieval and Data Governance

What to coverExplain how you would design a retrieval system that respects strict access control and data governance requirements. Address how you ensure that sensitive business data remains secure while being accessible to authorized agents.

Traditional ML vs. LLM Decision Framework

What to coverProvide a framework for deciding when to use a traditional machine learning model (e.g., gradient-boosted trees) versus an LLM for a specific business problem. Use examples like fraud detection or document classification to illustrate your reasoning.

Responsible AI and Guardrails

What to coverWhat specific techniques do you use to implement guardrails and groundedness checks in production? Discuss how you ensure compliance with data privacy regulations like PDPA while maintaining the utility of the AI system.

Cloud Infrastructure and Cost Management

What to coverDescribe your experience managing AI workloads on cloud platforms like AWS, Azure, or GCP. Explain how you balance performance and cost, specifically regarding GPU resource management and the use of batch processing.

End-to-End ML Lifecycle Management

What to coverDetail your process for taking an AI project from prototype to production. Focus on how you handle model upgrades, retirements, and the continuous monitoring of system reliability and latency.

Behavioural practice

Managing Stakeholder Expectations

What to coverDescribe a time you had to renegotiate the scope of an AI project with business stakeholders. How did you explain technical limitations without jargon and ensure the final deliverable met the business's core needs?

Cross-Team Project Coordination

What to coverGive an example of a project where you had to coordinate across Data Engineering and Enterprise Architecture teams. How did you manage milestones, report risks, and ensure alignment across different technical groups?

Mentoring and Technical Leadership

What to coverProvide an example of how you have mentored junior data scientists. How do you balance guiding their technical development with ensuring the project meets its delivery deadlines?

Problem Solving Under Uncertainty

What to coverDescribe a situation where an initial approach to an AI problem failed. How did you investigate the root cause, pivot your strategy, and eventually find a way forward to deliver measurable impact?

Sharpen your knowledge

Topics to prepare

LLM Evaluation and Monitoring Frameworks
The role requires building systems to measure groundedness, latency, and drift. Exercise: Design a testing pipeline for an LLM agent that includes automated evaluation of tool-call accuracy and safety guardrails. Relevance: Directly addresses the requirement to build evaluation and monitoring layers for production AI platforms.
Agentic Workflow Orchestration
Focus on multi-step agents, tool-calling, and human-escalation logic. Exercise: Create a flowchart for an agentic system that handles a complex business task, defining specific triggers for human intervention and policy compliance checks. Relevance: Essential for the responsibility of developing agentic workflows with strict policy adherence.
Responsible AI and Data Governance
Focus on PDPA compliance, access control, and audit trails. Exercise: Draft a data governance plan for a RAG system that ensures sensitive business knowledge is only accessible to authorized roles. Relevance: Required for designing secure knowledge retrieval systems within the airline's infrastructure.

Skills to revise

Production-Grade LLM Application Development
Hands-on implementation of retrieval-augmented generation (RAG) and model routing. Exercise: Build a prototype that routes queries between a high-performance model and a cost-effective model based on task complexity. Relevance: Directly supports the mandate to balance cost and performance in shared AI platforms.
Traditional Machine Learning Optimization
Application of gradient-boosted trees and anomaly detection. Exercise: Compare the performance of a gradient-boosted tree model versus an LLM for a specific classification task, documenting the trade-offs in latency and accuracy. Relevance: Required for the responsibility of deciding when a classic model outperforms an LLM for specific business use cases.
Cloud Infrastructure and Cost Management
Managing GPU resources and cloud deployment. Exercise: Create a cost-tracking dashboard or script that monitors token usage and cloud compute spend for a deployed model. Relevance: Aligns with the requirement to manage GPU resources and maintain cost-effective AI operations.

Your preparation checklist

Prepare a production-grade LLM project walkthrough
Select a past project where you deployed an LLM or agentic workflow to production. Prepare to discuss your methodology for structured evaluation, specifically how you measured groundedness, tool-call accuracy, and latency, and how you managed model drift or safety guardrails.
Develop a framework for model selection and cost optimization
Create a brief technical summary explaining your criteria for choosing between commercial LLMs (e.g., GPT, Claude) and open-weight models. Include how you would implement a routing strategy to balance performance, cost, and data privacy requirements in a secure cloud environment.
Draft a case study on traditional ML vs. LLM application
Prepare a concrete example where you chose a traditional machine learning technique (e.g., gradient-boosted trees for anomaly detection) over an LLM. Explain the business context, the performance metrics used, and why the simpler model provided superior value.
Formulate questions on AI governance and data privacy
Prepare thoughtful questions regarding how the AI Center of Excellence handles data governance and PDPA compliance when building shared platforms. Ask about their current approach to audit trails and access control for cross-departmental knowledge retrieval systems.
Outline a project management approach for cross-functional AI delivery
Prepare a brief explanation of how you manage stakeholder expectations when an AI project faces technical limitations. Focus on your process for defining success measures with business units and how you communicate risks and milestones to non-technical management.

Availability last checked: 18 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…