New opportunity

Research Engineer (Gen-AI Models)

National University of Singapore · Singapore

About this role

The Research Engineer will join A/P Wang Xinchao’s research group in the Department of Electrical and Computer Engineering at the National University of Singapore. Based at Kent Ridge Campus, the engineer will work closely with the Principal Investigator on research projects involving confidence- and similarity-based generative AI models from training through inference.

Responsibilities

Contribute to generative AI research and to on-chip ultra-low-power silicon systems using generative AI for real-time analytics at the edge.
Develop dialogue, speech signal processing and machine learning algorithms; contribute to an automatic speech recognition framework and related systems.
Advance research knowledge, publish in leading journals and conferences, and meet project deliverables and milestones.

Qualifications

A bachelor’s or master’s degree with an aptitude for the research is preferred. Prior related project experience is required; no specific number of years is stated.
Strong programming skills in at least one of C, C++, Matlab or Python.
Ability to work both independently and in a team, with strong organizational, writing and interpersonal communication skills. Candidates should be meticulous, able to work under pressure and committed to tight deadlines.
Candidates must be open to a fixed-term contract.

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

Skills for this role

Generative AIMachine LearningSpeech Signal ProcessingAutomatic Speech RecognitionCC++MatlabPythonResearch WritingCommunicationTeamwork

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

Generative AI research from training to inference
Contribute to confidence- and similarity-based generative AI models across training and inference, working closely with the Principal Investigator.
Real-time generative AI at the edge
Contribute to on-chip ultra-low-power silicon systems that use generative AI for real-time edge analytics.
Speech and dialogue systems
Develop dialogue, speech signal processing and machine learning algorithms, and contribute to an automatic speech recognition framework and related systems.
Research outputs and delivery
Advance research knowledge, publish in leading journals and conferences, and meet project deliverables and milestones.

Likely challenges to think through

Balancing model capability with edge constraints
A likely preparation challenge is explaining how model choices could affect power use and real-time performance. Prepare to discuss trade-offs and evaluation approaches without assuming a particular chip, model architecture or target metric.
Working across research areas
The advertised work spans generative AI, speech, recognition systems and silicon. Prepare to explain how you would investigate unfamiliar components and coordinate contributions; the posting says the engineer may work on one or more projects, not necessarily all areas in one system.
Balancing publication and deadlines
Research that supports a strong paper may compete for time with project milestones. Prepare to discuss how you would scope experiments, document results and communicate progress under tight deadlines.

Practice your answers

Technical practice

How would you use confidence and similarity in a generative AI model?

What to coverThe posting identifies confidence- and similarity-based Gen-AI models as a research area. Preparation suggestion: distinguish what each measure captures, where it might enter training or inference, and how you would test whether it improves results.

How would you evaluate a generative model from training through inference?

What to coverThe stated work spans training through inference. Preparation suggestion: explain how you would compare experimental results, check whether training gains persist at inference, and investigate a mismatch without assuming a particular model architecture.

What trade-offs arise when using generative AI in an ultra-low-power on-chip system?

What to coverThe role includes on-chip ultra-low-power silicon systems with Gen-AI. Preparation suggestion: discuss trade-offs among model capability, computation, memory, and power, while making clear which constraints would need measurement.

How would you assess whether an edge analytics system meets real-time needs?

What to coverReal-time analytics at the edge is an explicit project area. Preparation suggestion: describe how you would measure end-to-end latency alongside output quality and resource use, then diagnose which stage limits performance.

How would you approach a speech signal processing problem?

What to coverThe posting calls for contributions to speech signal processing algorithms. Preparation suggestion: explain how signal characteristics, noise, and preprocessing choices could affect downstream performance, and how you would evaluate a proposed change.

How would you contribute to implementing an automatic speech recognition framework?

What to coverImplementation of an automatic speech recognition framework and related systems is a stated duty. Preparation suggestion: outline how you would define component interfaces, test recognition behaviour, and trace errors through the processing pipeline without assuming a specified framework.

How would you evaluate a dialogue algorithm?

What to coverDialogue algorithms are named in the responsibilities. Preparation suggestion: describe how you would define the task, assess response quality and consistency, and examine failure cases; do not assume a particular application or dataset.

How would you turn a machine learning research idea into a reliable experiment?

What to coverThe role involves machine learning algorithms, prior project experience, and programming in at least one listed language. Preparation suggestion: discuss a baseline, controlled comparisons, implementation checks, and records that make results interpretable and reproducible.

Behavioural practice

When have you worked closely with a research lead on an uncertain problem?

What to coverThe posting says this engineer works closely with the PI. Preparation suggestion: use a real situation, explain how you aligned on goals and communicated findings, and state the result without implying a particular NUS working process.

When have you made progress on a project independently while remaining part of a team?

What to coverIndependent work and teamwork are both stated expectations. Preparation suggestion: give the situation, the decisions you owned, when you sought input, and the resulting contribution.

When have you managed a research or engineering milestone under a tight deadline?

What to coverThe posting mentions project deliverables, milestones, pressure, and tight deadlines. Preparation suggestion: describe the situation, how you prioritized and tracked work, what you communicated about risks, and the result.

When have you communicated technical findings in writing?

What to coverStrong writing is listed, and the role aims to publish in journals and conferences. Preparation suggestion: give a real example of the audience and purpose, how you supported your claims with evidence, and the outcome; it need not be a publication.

When has careful checking changed the outcome of your technical work?

What to coverThe qualifications call for meticulous work. Preparation suggestion: recount a specific issue you noticed, the checks you performed, how you addressed it, and what happened afterward.

Sharpen your knowledge

Topics to prepare

Confidence- and similarity-based Gen-AI models
The posting names these models from training through inference. As preparation, sketch how confidence and similarity could each influence a model’s output, then identify what you would check during training and inference. These approaches are a revision suggestion, not a specified project design.
Gen-AI for real-time edge analytics
The role includes on-chip ultra-low-power silicon systems using Gen-AI for real-time analytics at the edge. As preparation, explain the trade-offs among computation, power use and response time in a hypothetical edge deployment; the posting does not specify an architecture.
Dialogue, speech and machine learning algorithms
The posting explicitly calls for contributions across these areas. Revise by drawing a conceptual flow from speech input to dialogue output and identifying where signal processing and machine learning algorithms could contribute. Treat the flow as an interview exercise, not a stated system design.
Automatic speech recognition
The role includes implementing an automatic speech recognition framework and developing related systems. Revise the stages and possible sources of recognition error by tracing a short spoken utterance through a conceptual framework; no particular framework is named.

Skills to revise

Programming in a listed language
Strong programming in at least one of C, C++, Matlab or Python is specified. In one you know, write and test a small routine that compares numeric representations by similarity, then explain its inputs, edge cases and limitations. The routine is practice, not a prescribed implementation.
Reasoning through an ASR implementation
Because the job includes implementing an ASR framework, practise breaking a hypothetical recognition task into components, defining their inputs and outputs, and describing how you would isolate an error. No particular software or method is required by the posting.
Research writing
Publishing in leading journals and conferences and strong writing skills are stated expectations. Draft a one-page summary of a prior related project: question, approach, evidence, limitations and next step. Prior related project experience is required, but no specific project type is stated.
Milestone and team communication
The posting calls for project deliverables, tight deadlines, independent work and teamwork. Practise giving a brief project update that separates completed work, an unresolved issue, the next milestone and what input you need from the PI or team. This format is a preparation suggestion, not a stated NUS process.

Your preparation checklist

Prepare a related-project walkthrough
Prior related project experience is required. Select a project and prepare a concise account of your contribution, methods, results, limitations and what you would change.
Sketch a generative-model design
The role focuses on confidence- and similarity-based Gen-AI models from training to inference. As interview preparation, sketch how you might use confidence or similarity signals, then identify evaluation measures and failure cases; present this as a proposal, not a description of the group’s approach.
Work through an edge-deployment trade-off
The posting includes on-chip ultra-low-power systems for real-time edge analytics. Prepare a hypothetical design sketch showing how you would weigh model quality, latency, memory and power, and which measurements you would need.
Prepare a speech or dialogue example
The work includes dialogue and speech signal processing algorithms and implementation of an automatic speech recognition framework. Bring a relevant example, if you have one, or prepare an approach to an ASR component with tests and error-analysis steps.
Show programming and research-delivery evidence
Prepare code or a technical summary demonstrating one listed language—C, C++, Matlab or Python. Explain your decisions, how you checked the work, and how you organized milestones; the role also calls for writing, publication contributions and collaboration with the PI.
Draft questions for the research group
Ask which projects are the immediate priority, how model and system contributions are evaluated, what publication and delivery milestones are expected, and how the fixed-term contract is structured. The posting does not specify these details.

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…