Tech – Data Scientist: Responsible AI

Important

Interested in this project? When applying, remember to indicate the Project Title as listed above in the application form.

Role

Data Scientist

Division

AI Practice

Level

Undergraduate

Internship Duration

6 months, 12 months, 3 months

Project Description

Interested in this project?

Submit your application and indicate this project as your preference.

Apply Now!

Back to Data Scientist listings

As AI systems are increasingly deployed in real-world products and services, ensuring that they are safe, reliable, and responsible is critical. The Responsible AI team, part of GovTech's Artificial Intelligence Tech Practice, focuses on developing technical capabilities to define, measure, and mitigate Responsible AI risks within the Singapore public service. We aim to be at the forefront of the AI safety and responsible AI space through learning, experimentation, and applied research. We build practical frameworks and tools that translate Responsible AI principles into concrete engineering practices that support real deployment decisions. Some of our past works include RabakBench, KnowOrNot, MinorBench, and LionGuard. More details of our work can be found in our Responsible AI Playbook (https://go.gov.sg/responsibleaiplaybook) and division blog (https://blog.ai.gov.sg/). As this is a fast-moving space, you will be assigned to work with one of our team members on an applied research project relating to AI safety, robustness, fairness, interpretability, and/or explainability. You will be involved in all aspects of this process, from reviewing recent research to developing and deploying practical frameworks and tools, and contributing to knowledge sharing across the public sector, industry, and the open-source ecosystem. If you’re curious about how Responsible AI, practical engineering, and real-world policy come together and would like to contribute your ideas and skills in this space, you’ll find plenty of opportunities to learn, build, and make an impact with us.

Learning Outcomes

- Develop a strong understanding of key Responsible AI dimensions, such as safety, robustness, fairness, interpretability and/or explainability, in the context of public sector AI deployments. - Gain hands-on experience designing and building practical frameworks and tools for assessing and mitigating AI risks in real-world AI applications, while gaining technical experience in using and evaluating the latest LLMs. - Build applied research capabilities by translating emerging research ideas into practical solutions, and implementing them in a fast-paced, real-world environment. Build end-to-end data science and applied research and software engineering skills, including: - analyzing trace telemetry and execution logs - applying data-driven metrics to score non-deterministic trajectories - using real-world performance data to validate and refine theoretical eval frameworks - building intuitive user interfaces and APIs - designing scalable software system architectures for enterprise AI integration.

Prerequisites

- Python & Applied Data Science: Strong proficiency in Python with experience in data manipulation, API integrations (e.g., LLM SDKs), and practical machine learning fundamentals. - GenAI & Agentic Workflows: Practical experience or a solid understanding of LLM fundamentals, prompt engineering, and agentic concepts (e.g., multi-step reasoning trajectories, tool calling, and RAG). - Software Engineering Interest: Baseline experience or strong eagerness to learn full-stack software development (e.g., building Web UIs and REST APIs) to transform telemetry data into accessible software tools. - Problem Solving & Alignment: Ability to reason through ambiguous applied research problems independently, with a genuine interest in Responsible AI and public service impact. Nice-to-Haves (Bonus) - Eval & Observability Tooling: Familiarity with agentic evaluation frameworks (e.g., DeepEval) or trace observability platforms (e.g., Langfuse). - Engineering & MLOps: Exposure to containerization (Docker), cloud infrastructure (AWS/GCP/Azure), REST/OpenAPI standards, or ML deployment. Research Experience: Prior research background or experience translating academic AI/ML literature into practical code.