New opportunity

Research Associate (Embodied AI Safety)

Nanyang Technological University · Singapore

About this role

Nanyang Technological University’s School of Electrical and Electronic Engineering seeks a Research Associate to work on embodied AI safety at its main campus in Singapore. The role focuses on designing, developing and implementing novel algorithms and models to address emerging safety challenges, including attack and defense mechanisms.

Responsibilities

Collect and process data; conduct experiments and analyze results.
Assist with robot operation, debugging and system integration.
Provide general support for the lab.

Requirements

Master’s degree in Electrical Engineering, Computer Science or a related field.
Knowledge and experience in AI safety, computer vision and deep learning.
Proficiency in programming basics, such as Python.
Good written and oral communication skills, initiative and a strong drive to achieve objectives.

This is a full-time position. Only shortlisted candidates will be notified.

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

Skills for this role

AI safetyComputer visionDeep learningPythonData collectionData processingExperimental researchResults analysisRobot operationDebuggingSystem integrationWritten communicationOral communication

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

Develop embodied AI safety methods
The role centers on designing, developing and implementing novel algorithms and models for embodied AI safety, including attack and defense mechanisms. The posting calls for knowledge and experience in AI safety, computer vision and deep learning.
Run the research workflow
Collect and process data, conduct experiments and analyze results. Programming basics such as Python are a stated requirement.
Support robot-based implementation
Assist with robot operation, debugging and system integration so research methods can be tested beyond model development alone.
Contribute to the lab’s objectives
Provide general lab support, communicate well in writing and orally, and show initiative in pursuing objectives.

Likely challenges to think through

Evaluating attacks without losing sight of defenses
Preparation hypothesis: An attack may expose a weakness without showing that a defense addresses it. Prepare to explain how you would compare attack and defense results, including limitations, without assuming a particular threat model or benchmark.
Drawing sound conclusions from experimental data
Preparation hypothesis: Data collection and processing choices could affect results, making apparent improvements hard to interpret. Prepare an example of how you would document processing decisions, check results and communicate uncertainty.
Moving from a model to a working robot system
Preparation hypothesis: A promising algorithm may be difficult to integrate or debug during robot operation. Prepare to describe a stepwise approach to isolating model, software and integration issues while balancing research progress with lab support duties.

Practice your answers

Technical practice

How would you define a safety problem for an embodied AI system?

What to coverThe posting focuses on emerging embodied AI safety challenges. Suggested preparation: describe the system’s observations and actions, a specific unsafe outcome, and assumptions you would test before designing a model.

How would you study an attack against an embodied AI model?

What to coverThe role explicitly includes attack mechanisms. Suggested preparation: define what an attacker could change, establish a baseline, and measure both attack success and any effect on normal operation.

How would you design and evaluate a defense?

What to coverThe posting calls for developing algorithms and models, including defenses. Suggested preparation: explain the defense’s mechanism and compare safety gains against trade-offs such as accuracy or computational cost.

What computer-vision failures could matter for robot safety?

What to coverComputer vision is a stated knowledge area, and the role includes assisting with robot operation. Suggested preparation: discuss how changes in visual inputs could affect predictions, how you would detect failures, and why their consequences depend on the robot’s actions.

How would you choose and develop a deep-learning model for this research?

What to coverThe posting names deep learning and novel model development but specifies no architecture. Suggested preparation: explain how you would select a baseline, identify the proposed change, and test whether it improves the safety objective rather than only average performance.

How would you collect and process data for a safety experiment?

What to coverData collection and processing are explicit responsibilities. Suggested preparation: cover data quality, consistent labels or records, and how you would avoid losing uncommon but important failure cases during processing.

How would you structure an experiment to compare two safety approaches?

What to coverThe role includes conducting experiments. Suggested preparation: identify a baseline, controlled conditions, relevant safety and task-performance measures, and what would make the comparison reproducible.

How would you interpret mixed experimental results?

What to coverAnalyzing results is a stated responsibility. Suggested preparation: explain how you would examine variation across conditions, distinguish a promising result from a fragile one, and decide what follow-up experiment is warranted.

How would you use Python to make research experiments easier to reproduce?

What to coverThe posting asks for proficiency in programming basics such as Python. Suggested preparation: discuss clear data-processing code, recorded configurations, and checks that catch errors before results are compared; no particular framework is specified.

How would you troubleshoot a model-to-robot integration problem?

What to coverThe role includes assisting with robot operation, debugging, and system integration. Suggested preparation: trace inputs and outputs across components, isolate whether the issue arises in data, model behavior, or integration, and verify a fix before further operation.

Behavioural practice

Tell me about a time you took initiative on an unclear research or technical task.

What to coverThe posting explicitly values initiative and drive to achieve objectives. Suggested preparation: give a real situation, the steps you chose without assuming direction was provided, and the result or lesson.

Describe a time you supported a team while managing your own work.

What to coverProviding general lab support is a stated responsibility. Suggested preparation: use a real situation to explain what support was needed, how you prioritized and coordinated your actions, and the outcome.

Tell me about a time you communicated a complex result to others.

What to coverThe posting asks for good written and oral communication and includes results analysis. Suggested preparation: describe the audience and situation, how you explained the evidence and its limits, and what resulted.

Describe a time debugging changed your initial understanding of a problem.

What to coverDebugging and system integration are part of the advertised work. Suggested preparation: give a real example showing the initial symptom, the checks you performed, what you changed, and the verified result.

Sharpen your knowledge

Topics to prepare

Embodied AI safety and threat models
Suggested exercise: Sketch how an attack on a robot's perception could affect its behavior, then compare two possible defenses and their limitations. Relevant because the role focuses on embodied AI safety algorithms, including attack and defense mechanisms; the posting does not specify particular attacks.
Computer vision and deep learning failure modes
Suggested exercise: Explain how a vision model processes input and where errors could arise under changed visual conditions. Relevant because knowledge and experience in computer vision and deep learning are stated requirements.
Experimental design and interpretation
Suggested exercise: Define a comparison between a proposed defense and an unmodified model, including what data, measurements and limitations you would report. Relevant because the role includes conducting experiments and analyzing results.
Robot-system interfaces
Suggested exercise: Draw a conceptual path from sensor input through model output to robot action, and identify where faults could be isolated. Relevant because the role includes assisting with robot operation, debugging and system integration; no robot platform is specified.

Skills to revise

Python data preparation
Suggested exercise: Write a short Python program to check a sample dataset for missing records, standardize its format and summarize its contents. Relevant because the posting names data collection and processing and proficiency in programming basics such as Python.
Model implementation
Suggested exercise: Implement a small model change and verify that its inputs, outputs and results are recorded consistently. Relevant because the role focuses on designing, developing and implementing algorithms and models; no framework is specified.
Experiment execution and results analysis
Suggested exercise: Run a controlled comparison on sample data, tabulate the results and identify one plausible limitation. Relevant because conducting experiments and analyzing results are explicit responsibilities.
Robot debugging and integration
Suggested exercise: Given a hypothetical mismatch between a model output and a robot action, trace the data passed between components and propose checks to isolate the fault. Relevant because the role includes assisting with robot operation, debugging and system integration; this is preparation, not a stated platform requirement.
Research communication
Suggested exercise: Prepare a brief written results summary and a two-minute oral explanation covering the method, findings and uncertainty. Relevant because good written and oral communication skills are stated requirements.

Your preparation checklist

Map the stated requirements to evidence
Prepare a one-page list of the master’s degree requirement and examples, if available, of AI safety, computer vision, deep learning, Python, communication and initiative. Note where an example is missing rather than implying experience.
Build a project walkthrough
Choose a relevant project, if available, and prepare a five-minute account of the problem, your contribution, model or algorithm choices, experiments, results and limitations. Connect it to embodied AI safety only where the connection is genuine.
Sketch an attack-and-defense design exercise
As interview preparation, outline a hypothetical safety challenge for a vision-enabled robot, a possible attack and defense, and how you would test the defense. The posting names attack and defense mechanisms but does not prescribe a specific approach.
Prepare an experiment record
Bring a concise example, if available, showing how you collected or processed data, controlled an experiment and analyzed its results. Include a metric, a comparison and what the results did not establish.
Rehearse a Python explanation
Practice explaining or writing a small Python routine for data processing or result analysis, including how you would check its output. The posting asks for programming basics such as Python, not a named framework.
Outline a robot troubleshooting approach
Prepare a short, clearly hypothetical walkthrough of how you would isolate a robot-operation or system-integration fault, record observations and verify a fix. The role includes assisting with operation, debugging and integration.
Draft questions and a concise research update
Prepare a brief spoken or written update on an example of self-directed work, plus questions about the lab’s safety challenges, available experimental setup and how research and lab-support tasks are balanced. These are questions to ask, not details established 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…