New opportunity

AI Engineer 2 (Data & AI Department)

EMA Energy Market Authority · Singapore

About this role

The AI Engineer 2 develops and operationalises production-ready AI solutions for the Energy Market Authority’s energy-sector operations and regulatory functions. The role spans AI/ML model development and deployment, MLOps lifecycle management, and technical leadership across business and technical teams. The posting lists the location as EMA HQ, Singapore, and the role as full-time.

Responsibilities

Collaborate with data scientists and data engineers to design, test, implement and deploy AI/ML models across the full lifecycle, delivering reliable and scalable solutions for EMA use cases.
Establish MLOps practices from code commit through model retirement. Build continuous integration, deployment and training (CI/CD/CT) pipelines; manage version control for code, data and models; and implement automated retraining, performance monitoring, data drift detection and guardrails for critical infrastructure environments.
Engage senior management, business users, data scientists, and vendor and IT teams. Explain AI outputs to non-technical audiences and balance business outcomes, system performance constraints and compliance requirements.

Requirements

Higher tertiary background in computer science, mathematics, physics, data science or a related quantitative field; specialised AI certifications are advantageous.
Minimum 5–10 years of professional experience developing, deploying and maintaining enterprise-grade AI/ML models in production, with demonstrable experience across the full MLOps lifecycle.
Expert knowledge of supervised and unsupervised learning, deep learning and reinforcement learning; advanced Python skills and experience with frameworks such as PyTorch, TensorFlow and Pandas.
Knowledge of automated model training, testing, deployment and monitoring, plus experience with model optimisation techniques including quantisation, pruning and knowledge distillation.
Strong stakeholder management, interpersonal and communication skills; systems thinking, independent working ability and sound judgement about technical trade-offs.

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

Skills for this role

AI/ML modellingMLOpsSupervised learningUnsupervised learningDeep learningReinforcement learningPythonPyTorchTensorFlowPandasCI/CD/CTVersion controlAutomated model retrainingModel monitoringData drift detectionModel governanceQuantisationPruningKnowledge distillationStakeholder managementTechnical communicationSystems thinking

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-ready AI solutions
The role leads the design, testing and deployment of scalable AI/ML models with data scientists and data engineers for EMA’s energy-sector operations and regulatory functions.
End-to-end MLOps
The posting calls for MLOps foundations from code commit to model retirement, including CI/CD/CT pipelines and version control for code, data and models.
Model reliability and safety
Automated retraining, performance monitoring, drift detection, guardrails and model governance are central to keeping models effective in critical infrastructure environments.
Technical leadership across stakeholders
The engineer aligns senior management, business users, data scientists and vendor/IT teams, translating AI outputs into actionable insights while weighing business outcomes, system limits and compliance.
Deployment-aware model optimisation
The requirements include experience with quantisation, pruning and knowledge distillation, alongside production experience with enterprise-grade models.

Likely challenges to think through

Automating change without weakening safeguards
Preparation hypothesis: continuous training and automated retraining may need to be balanced against reliability and safety. Prepare an example of how you would detect drift, validate a replacement model, apply guardrails and decide whether to deploy or hold it.
Making model trade-offs under system constraints
Preparation hypothesis: a more capable model may be harder to deploy reliably or promptly. Be ready to explain how you would compare model performance with deployment constraints and when optimisation techniques would be worth using.
Aligning different definitions of success
Preparation hypothesis: business value, technical feasibility and compliance may pull stakeholders in different directions. Prepare a concise example of how you would explain an AI output to non-technical decision-makers, surface the trade-off and reach a decision with technical and business teams.

Practice your answers

Technical practice

How would you choose among supervised, unsupervised, deep, and reinforcement learning for an AI use case?

What to coverThe posting names all four approaches but does not specify a use case. Suggested preparation: explain how the available data, decision objective, evaluation method, and operational risk would guide your choice.

What would you test before releasing an AI/ML model for a critical energy-sector function?

What to coverThe role calls for production-ready, reliable, scalable solutions. Suggested preparation: distinguish model-quality tests from integration, performance, and safety checks, and explain what evidence would justify release.

How would you structure maintainable Python code for model development and deployment?

What to coverThe posting seeks advanced Python skills and names PyTorch, TensorFlow, and Pandas. Suggested preparation: discuss reproducible data preparation, separation of training and inference code, and tests without assuming a particular framework is mandated.

How would you design CI, deployment, and continuous-training pipelines across the model lifecycle?

What to coverThe role explicitly calls for CI/CD/CT from code commit onward. Suggested preparation: describe pipeline stages, automated checks, promotion decisions, and how training changes reach production safely.

How would you keep code, data, and model versions traceable through model retirement?

What to coverThe posting specifies version control across all three and an end-to-end lifecycle through retirement. Suggested preparation: explain how to reproduce a release, identify its inputs, manage changes, and retire an obsolete model.

How would you detect and investigate data drift and declining model performance?

What to coverPerformance monitoring and data drift detection are stated responsibilities. Suggested preparation: distinguish changes in input data from changes in model outcomes, and describe how you would verify an alert before choosing a response.

When should automated retraining run, and what should prevent an unsafe update?

What to coverThe role includes automated retraining infrastructure and guardrails. Suggested preparation: discuss retraining triggers, validation against the current model, and conditions for withholding or reverting a release.

What model guardrails and governance would you apply in a critical-infrastructure setting?

What to coverThe posting links model governance and guardrails to reliability and safety. Suggested preparation: explain how you would define acceptable operating conditions, monitor exceptions, and make model decisions reviewable without assuming a specific policy.

How would you compare quantisation, pruning, and knowledge distillation?

What to coverAll three optimisation techniques appear in the requirements. Suggested preparation: explain how each may affect resource use, inference performance, and model quality, and how you would test whether an optimisation is acceptable.

How would you troubleshoot a model that performs well in testing but poorly after deployment?

What to coverThe role spans deployment, monitoring, and maintenance of production models. Suggested preparation: trace possible differences in data, preprocessing, inference behaviour, and system constraints before deciding whether to fix the pipeline, update the model, or roll back.

Behavioural practice

Tell me about a time you worked with data scientists and data engineers to deliver a production model.

What to coverThe role explicitly requires this collaboration across the AI/ML lifecycle. Suggested preparation: give the situation, your actions to coordinate design and deployment, and the result, including any reliability or delivery evidence you have.

Describe a time you explained a complex AI output to a non-technical audience.

What to coverThe posting calls for translating AI outputs into actionable insights for business users and senior management. Suggested preparation: describe the audience and decision, how you conveyed limitations as well as findings, and what resulted.

Give an example of resolving a trade-off between a business objective, system performance, and compliance.

What to coverBalancing these considerations is a stated responsibility. Suggested preparation: set out the competing needs, how you reached and communicated a decision, and its outcome without implying that every objective was fully met.

Tell me about coordinating with vendor or IT teams on a technical delivery issue.

What to coverThe role includes engagement with vendor and IT teams. Suggested preparation: describe the issue, how you clarified responsibilities and resolved dependencies, and the resulting effect on delivery.

Describe a time you exercised independent judgement under complex technical constraints.

What to coverThe posting seeks independent working ability and systems thinking. Suggested preparation: explain the situation, the wider system effects you considered, the action you took or escalated, and the result.

Sharpen your knowledge

Topics to prepare

End-to-end MLOps lifecycle
Posting: The role covers code commit through model retirement, including CI/CD/CT and version control for code, data and models. Suggested exercise: Sketch a lifecycle diagram and explain the checks and records needed at each transition.
Monitoring, retraining and model governance
Posting: Automated retraining, performance monitoring, data drift detection and guardrails support reliability and safety in critical infrastructure. Suggested exercise: Reason through a hypothetical drift alert: distinguish drift from declining performance and decide when to investigate, retrain or hold deployment.
AI/ML modelling choices
Posting: Expert proficiency is sought in supervised and unsupervised learning, deep learning and reinforcement learning. Suggested exercise: For a hypothetical EMA use case, compare when each approach would be appropriate and what evidence would justify the choice; the posting does not specify a particular use case.
Model optimisation trade-offs
Posting: Experience with quantisation, pruning and knowledge distillation is requested. Suggested exercise: Compare how each technique might affect model quality, resource use and production constraints, and identify what you would measure before adoption.
Systems thinking and compliance trade-offs
Posting: Technical leadership involves balancing business outcomes, system performance limitations and compliance requirements. Suggested exercise: Work through a hypothetical deployment decision and explain how reliability, safety and stakeholder needs would affect the recommendation.

Skills to revise

Build and test models in Python
Posting: Advanced Python skills and experience with frameworks such as PyTorch, TensorFlow and Pandas are requested. Suggested exercise: Use Python and a listed framework to prepare data, train and test a small model, then explain the design and evaluation choices.
Design a CI/CD/CT workflow
Posting: The role designs pipelines for integration, deployment and continuous training. Suggested exercise: Draft a pipeline with automated training and testing, deployment checks, and versioned code, data and models; explain what would block a release.
Specify monitoring and retraining controls
Posting: The role implements monitoring, drift detection, automated retraining and guardrails. Suggested exercise: Write a monitoring-and-response plan for a hypothetical production model, including alert conditions, retraining checks and a safe response to failed checks.
Evaluate an optimisation change
Posting: Quantisation, pruning and knowledge distillation are named techniques. Suggested exercise: Apply one to a small model using an appropriate listed framework, compare results before and after, and state whether the trade-off supports deployment.
Explain a technical decision to stakeholders
Posting: The role engages senior management, business users, data scientists and vendor/IT teams and translates AI outputs for non-technical audiences. Suggested exercise: Deliver a short spoken briefing on a hypothetical model decision, separating the business implication, technical limitation and compliance question.

Your preparation checklist

Map evidence to the stated requirements
Prepare a one-page list of examples and supporting artefacts for production AI/ML experience, full-lifecycle MLOps, Python, modelling, optimisation and stakeholder leadership. Note that the posting describes specialised AI certifications as advantageous, not required.
Build a production-model walkthrough
Choose a relevant project you can discuss and outline the problem, model choice, testing, deployment, maintenance, measurable results and your decisions. The role calls for reliable, scalable AI solutions for energy-sector operations and regulatory functions; do not assume a specific EMA use case.
Sketch an end-to-end MLOps design
Prepare a diagram from code commit to model retirement covering code, data and model versioning; CI/CD/CT; automated retraining; performance and drift monitoring; and governance and safety guardrails. Be ready to explain how you would test the design for a critical-infrastructure setting.
Rehearse modelling and coding trade-offs
Prepare concise examples of when supervised, unsupervised, deep or reinforcement learning is appropriate, including when it is not. Select Python work using relevant tools such as Pandas, PyTorch or TensorFlow that you can explain in detail.
Quantify an optimisation decision
If you have a suitable example, compare accuracy, latency and resource costs before and after quantisation, pruning or knowledge distillation. Explain the constraint that drove the choice and how you checked its effect on model quality.
Practise a stakeholder briefing
Draft a short, non-technical explanation of an AI result and a decision it supports. Rehearse how you would discuss business outcomes, system limitations and compliance trade-offs with senior management, business users, data scientists and vendor or IT teams.
Prepare questions about the operating context
Ask which energy-sector use cases are priorities, what reliability and compliance constraints apply, how model governance works, and how responsibilities are divided among data scientists, data engineers and vendor or IT teams. These are questions to resolve, not details specified by the posting.

Availability last checked: 4 Nov 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…