Research Fellow (Computer Science/Computer Engineering/Electronics Engineering/AI/Machine Learning)
Nanyang Technological University · Singapore
About this role
The College of Computing and Data Science (CCDS) at Nanyang Technological University (NTU) is seeking a Research Fellow to join their team in Singapore. This role focuses on advancing cutting-edge research in AI and computing. Key Responsibilities: Conduct research on 3D/4D scene reconstruction and rendering, including NeRF, 3D Gaussian Splatting, and novel view synthesis, as well as scene understanding and interaction. Develop generative AI models for video and 3D/4D generation, world models, and vision-language models (VLMs) or vision-language-action (VLA) systems, supported by large-scale data pipelines. Additional duties include publishing in top-tier venues, contributing to open-source releases, mentoring junior researchers, and reporting project progress. Job Requirements: Candidates must hold a PhD degree in Computer Science, Computer Engineering, Electronics Engineering, AI, Machine Learning, or an equivalent field. A strong research record in computer vision or machine learning is required, along with solid programming skills in Python and PyTorch. Applicants should demonstrate an ability to conduct innovative research and possess strong analytical, communication, and teamwork skills. The position is based at the NTU Main Campus.
For full details about this role, click “Apply on company site” to view the employer’s original posting.
A focused preparation guide, built around this job’s responsibilities and requirements.
✦ AI-generated guide
Preparation suggestions, not the employer’s actual interview questions. Always check the original posting for current requirements.
01
Understand the role
What matters most
Advancement of 3D/4D Scene Reconstruction
The role requires active research and development in complex rendering techniques such as NeRF, 3D Gaussian Splatting, and novel view synthesis. Success depends on the ability to push the boundaries of current scene understanding and interaction capabilities.
Development of Generative AI and VLM/VLA Systems
The candidate must build and refine generative models for video and 3D/4D generation, alongside vision-language-action systems. This necessitates the design and management of large-scale data pipelines to support model training and performance.
Academic Output and Mentorship
Beyond technical development, the fellow is expected to publish findings in top-tier venues, contribute to open-source initiatives, and provide mentorship to junior researchers, ensuring the research has both academic impact and practical utility.
Likely challenges to think through
Balancing Innovation with Publication Pressure
The requirement to conduct highly innovative research while simultaneously meeting the demands of top-tier publication venues creates a high-pressure environment. Prepare by demonstrating a track record of managing long-term research projects that yield high-quality, peer-reviewed results.
Complexity of Large-Scale Data Pipelines
Developing VLM/VLA systems requires robust data engineering. The challenge lies in scaling these pipelines effectively for complex 3D/4D data. Prepare by highlighting your experience in optimizing data workflows and handling large datasets within PyTorch environments.
03
Practice your answers
Technical practice
NeRF vs. 3D Gaussian Splatting
What to coverCompare the computational trade-offs between Neural Radiance Fields (NeRF) and 3D Gaussian Splatting for real-time rendering. Discuss how each approach handles scene representation, training time, and rendering quality in the context of novel view synthesis.
Scalable Data Pipelines for Generative AI
What to coverDescribe the architectural considerations for building large-scale data pipelines to support training generative models for 3D/4D video. Focus on data ingestion, preprocessing, and storage strategies necessary for high-fidelity vision-language-action (VLA) training.
Vision-Language-Action (VLA) Model Integration
What to coverExplain the challenges of aligning visual perception with robotic or agentic actions in VLA systems. Discuss how to handle temporal consistency and multi-modal data fusion when training these models.
World Model Architectures
What to coverDiscuss the role of world models in generative AI for 3D/4D environments. Explain how these models predict future states and the importance of latent space representations for scene understanding.
PyTorch Optimization for Research
What to coverDetail your approach to optimizing PyTorch code for high-performance research tasks. Suggestion: Be prepared to discuss techniques like mixed-precision training, distributed data parallel (DDP) strategies, and custom CUDA kernels if relevant to your past work.
Scene Understanding and Interaction
What to coverHow do you approach the problem of semantic scene understanding in 3D reconstructions? Discuss methods for segmenting and interacting with objects within a reconstructed scene.
Evaluating Generative 3D/4D Models
What to coverWhat metrics and qualitative evaluation methods do you prioritize when assessing the performance of generative 3D/4D models? Address the difficulty of measuring temporal coherence and visual fidelity in novel view synthesis.
Handling Sparse Data in 3D Reconstruction
What to coverExplain strategies for improving 3D scene reconstruction quality when input data is sparse or noisy. Discuss the trade-offs between using priors versus relying solely on the input imagery.
Behavioural practice
Mentoring Junior Researchers
What to coverDescribe a situation where you mentored a junior researcher or student. Focus on how you balanced their learning needs with the project's technical requirements and deadlines.
Collaborative Research and Teamwork
What to coverProvide an example of a time you worked in a diverse research team to solve a complex problem. Highlight your role in facilitating communication and ensuring the team reached a consensus on the technical approach.
Navigating Research Deadlines and Publication
What to coverTell me about a time you had to manage the pressure of preparing a paper for a top-tier venue while simultaneously handling ongoing project reporting. How did you prioritize your tasks to ensure quality?
Proactive Problem Solving in Research
What to coverDescribe a time when your initial research hypothesis failed or a model did not perform as expected. How did you pivot your strategy, and what steps did you take to identify the root cause?
02
Sharpen your knowledge
Topics to prepare
3D/4D Scene Reconstruction and Rendering
Review the theoretical foundations of Neural Radiance Fields (NeRF) and 3D Gaussian Splatting. Exercise: Summarize the mathematical differences between volumetric rendering and point-based splatting to prepare for discussing novel view synthesis techniques in an academic interview.
Generative AI and World Models
Study the architecture of modern generative models for video and 3D generation. Exercise: Draft a brief research proposal or critique on how world models can be integrated with Vision-Language-Action (VLA) systems to improve robotic or agentic interaction.
Skills to revise
Advanced PyTorch Implementation
Practice implementing custom loss functions and training loops for complex 3D vision tasks. Exercise: Build a small-scale prototype using PyTorch that demonstrates efficient handling of large-scale data pipelines for 3D point cloud or video data.
Research Communication and Mentorship
Prepare to discuss your track record of publishing in top-tier venues and your approach to guiding junior researchers. Exercise: Prepare a concise summary of your most significant research contribution, focusing on the innovation, methodology, and impact, suitable for a peer-level technical discussion.
04
Your preparation checklist
Synthesize your research portfolio
Prepare a concise summary of your PhD research and subsequent work, specifically highlighting publications in top-tier computer vision or machine learning venues. Ensure you can clearly articulate how your past projects align with 3D/4D scene reconstruction, NeRF, or generative AI.
Demonstrate technical proficiency in PyTorch
Curate a selection of your open-source contributions or code repositories that showcase your ability to build complex models in Python and PyTorch. Be ready to discuss your experience with large-scale data pipelines and the implementation of novel architectures.
Prepare a research vision statement
Draft a brief proposal or discussion points regarding how you would approach current challenges in 3D/4D generation or vision-language-action (VLA) systems. This demonstrates your ability to conduct innovative, independent research.
Formulate questions for the research team
Prepare thoughtful questions regarding the current research focus of the College of Computing and Data Science, such as the specific hardware infrastructure available for large-scale training or the team's current approach to mentoring junior researchers.
Availability last checked: 18 Oct 2026. Status can change after a check; confirm on the employer’s website. Report an issue · Our checking policy
Share this opportunityKnow someone who would be a great fit?