New opportunity

Research Engineer II (Trust Technologies)

Nanyang Technological University · Singapore

About this role

The National Centre for Research in Digital Trust (DTC) at Nanyang Technological University (NTU) is seeking a hands-on Research Engineer to build and implement AI components for Projectx. This execution-focused role supports a national program aimed at establishing Singapore as a strong trust hub by creating a trusted digital environment. Responsibilities include end-to-end development of local LLM systems, RAG pipelines, and trustworthy AI controls. Key Responsibilities: Build end-to-end local LLM/RAG pipelines for conversational analysis of large unstructured datasets. Fine-tune and adapt LLM/NLP models using supervised fine-tuning and LoRA/PEFT techniques, benchmarking them against baselines. Implement NLP capabilities such as entity extraction, resolution, and relationship analysis. Develop backend APIs, database/vector-store integration, local model serving, GPU inference, and containerized deployment. Implement access control, audit logging, traceable retrieval/output provenance, model isolation, and human-review workflows. Conduct systematic evaluation, security/robustness testing, error analysis, performance optimization, and regression testing. Maintain reproducible code, versioning, experiment records, and technical documentation. Communicate results to technical and non-technical audiences. Job Requirements: Bachelor’s degree in computer science, AI, engineering, data science, software engineering, or a related field. Minimum 4 years of relevant working experience. Strong Python skills with hands-on experience in LLM/NLP development. Practical experience with RAG and model fine-tuning/PEFT. Working knowledge of Linux, Git, Docker, SQL, APIs, and modern data/model-serving workflows. Preferred Qualifications: Experience with PyTorch and/or Hugging Face. Experience with local GPU deployment, vector databases, NLP analytics, or trustworthy AI/security controls is an advantage. Candidates must be independent, execution-oriented, and methodical.

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

Skills for this role

PythonLLMNLPRAGFine-tuningPEFTPyTorchHugging FaceLinuxGitDockerSQLAPIsGPU inferenceVector databases

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

End-to-End Local LLM and RAG Pipeline Development
The core requirement is to build and deploy local AI systems capable of processing large unstructured datasets. This involves integrating vector databases, managing GPU inference, and ensuring the entire pipeline is containerized and functional within controlled or on-premise environments.
Implementation of Trustworthy AI Controls
As part of a national digital trust program, the role requires embedding security and governance into the AI lifecycle. This includes developing access controls, audit logging, traceable retrieval provenance, and model isolation to ensure outputs are reliable and verifiable.
Model Adaptation and Systematic Evaluation
The engineer must perform supervised fine-tuning and utilize PEFT/LoRA techniques to adapt models. Success is measured by rigorous benchmarking against baselines, systematic error analysis, and performance optimization to ensure the models meet project-specific accuracy and robustness standards.

Likely challenges to think through

Balancing Performance with Local Resource Constraints
Deploying LLMs in local or on-premise environments often involves strict hardware limitations. Prepare by demonstrating experience in optimizing model serving, managing GPU memory, and selecting efficient quantization or inference strategies that maintain performance without exceeding local compute capacity.
Maintaining Reproducibility in Research-to-Production Transitions
The role demands methodical documentation and version control for both code and model configurations. Prepare by highlighting your experience with experiment tracking tools and structured workflows that ensure research prototypes can be reliably transitioned into stable, production-ready components.

Practice your answers

Technical practice

Designing Local RAG Pipelines

What to coverExplain your approach to building a RAG pipeline for large unstructured datasets in a local, on-premise environment. Focus on the trade-offs between different vector database choices, document chunking strategies, and retrieval accuracy optimization.

Fine-tuning with PEFT and LoRA

What to coverDescribe the process of adapting a pre-trained LLM using LoRA or other PEFT techniques. Discuss how you manage memory constraints during training and how you validate that the fine-tuned model maintains performance against original baselines.

Local GPU Inference and Deployment

What to coverHow do you optimize a model for local GPU inference? Address considerations such as model quantization, batching strategies, and containerization using Docker to ensure reliable performance in a controlled environment.

Implementing Trustworthy AI Controls

What to coverThe role emphasizes trustworthy AI. Describe how you would implement traceable retrieval, output provenance, and audit logging within an LLM system to ensure transparency and accountability.

NLP Analytics and Entity Resolution

What to coverDiscuss your experience with entity extraction and resolution in large datasets. Suggest how you handle ambiguity in entity matching and how you integrate these NLP capabilities into a broader backend API architecture.

Systematic Evaluation and Robustness

What to coverHow do you conduct security and robustness testing for LLM-based applications? Explain your methodology for error analysis and how you use regression testing to prevent performance degradation after model updates.

Data Pipeline and Database Integration

What to coverDescribe your experience integrating SQL and vector databases into an end-to-end AI pipeline. Focus on how you ensure data consistency and efficient retrieval for real-time conversational analysis.

Reproducibility and Experiment Tracking

What to coverGiven the research-focused nature of this role, how do you maintain reproducible code and experiment records? Suggest tools or workflows for versioning models, configurations, and datasets.

Behavioural practice

Delivering Under Tight Timelines

What to coverDescribe a situation where you had to deliver a complex technical component under a tight deadline. Focus on how you prioritized tasks, managed scope, and ensured the final output met quality standards.

Communicating Technical Results

What to coverProvide an example of when you had to explain a complex AI model or technical finding to a non-technical stakeholder. How did you adapt your communication style to ensure they understood the implications of your work?

Methodical Problem Solving

What to coverTell me about a time you encountered a persistent error in a model or pipeline. Walk through your methodical approach to debugging, isolating the root cause, and implementing a permanent fix.

Working Independently in Research

What to coverDescribe a project where you were required to work independently to translate high-level requirements into a working prototype. How did you manage your own progress and ensure alignment with project goals?

Sharpen your knowledge

Topics to prepare

Trustworthy AI and Security Controls
The role requires implementing access control, audit logging, and output provenance. Review frameworks for AI security and model isolation to ensure you can discuss how to maintain data integrity and system security in a local, on-premise environment.
RAG Pipeline Evaluation and Benchmarking
The position emphasizes systematic evaluation and benchmarking against baselines. Study methodologies for evaluating retrieval accuracy and generation quality in RAG systems, including how to perform error analysis and regression testing on unstructured datasets.

Skills to revise

PEFT and LoRA Fine-Tuning
Practice fine-tuning local LLMs using Parameter-Efficient Fine-Tuning (PEFT) techniques like LoRA. Use a library like Hugging Face PEFT to adapt a model to a specific task and document the performance metrics compared to a base model.
Local GPU Deployment and Containerization
Develop a workflow to containerize a local LLM application using Docker. Ensure you can manage GPU inference within a containerized environment, as the role requires deploying models in controlled or on-premise settings.
Vector Database Integration
Build a small-scale RAG prototype that integrates a vector database with a local LLM. Focus on the end-to-end flow from data ingestion and embedding generation to retrieval and final output generation.

Your preparation checklist

Prepare a technical project portfolio
Compile documentation and code samples demonstrating end-to-end RAG pipelines and LLM fine-tuning using LoRA/PEFT. Ensure your portfolio highlights reproducible experiments, version control practices, and experience with local GPU deployment.
Review trustworthy AI and security controls
Prepare to discuss how you implement audit logging, access control, and output provenance in LLM systems. Be ready to explain your approach to model isolation and human-in-the-loop workflows for sensitive data analysis.
Demonstrate deployment and infrastructure proficiency
Be prepared to explain your experience with containerization (Docker), Linux environments, and API development. Focus on how you manage model serving and vector database integration for on-premise or controlled environments.
Formulate questions regarding Projectx
Prepare thoughtful questions about the specific unstructured datasets involved in Projectx, the current evaluation metrics for model robustness, and the balance between performance optimization and security requirements in the team's workflow.
Practice technical communication
Prepare to explain complex NLP concepts like entity resolution or model benchmarking to both technical peers and non-technical stakeholders, as this is a core requirement for the role.

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…