Tech – Data Scientist: AI Evaluation of MOE DXD Products

Important

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

Role

Data Scientist

Division

Central Administration & Social

Level

Undergraduate

Internship Duration

12 months, 6 months

Project Description

Interested in this project?

Submit your application and indicate this project as your preference.

Apply Now!

Back to Data Scientist listings

The MOE AI Programmes builds and deploys AI products for the education sector. AI Evaluation is a practice that establishes how these AI products are measured before and after they are deployed to schools, teachers and students. It covers the design of evaluation framework, the construction of golden set, and the running of LLM-as-judge and human-in-the-loop calibration.

Learning Outcomes

- AI Evaluation Design: Learn to translate product goals into measurable evaluation criteria, rubrics, and test cases for generative and computer vision AI systems deployed in real school settings. - LLM-as-Judge and Automated Evaluation: Develop practical skills in building automated evaluation pipelines, benchmarking model and prompt versions, and validating automated judgments against human ground truth. - Responsible AI in the Public Sector: Build familiarity with AI governance and risk frameworks, and learn how evaluation evidence informs go/no-go decisions and ongoing monitoring for public-facing AI products.

Prerequisites

- Experience with AI coding assistants / agentic CLI tools (e.g. Claude Code, Gemini CLI, Codex), LLM APIs, or evaluation frameworks will be an advantage - Working knowledge of statistics — sampling, hypothesis testing, and interpreting metrics such as precision, recall, and inter-rater agreement — with an interest in designing evaluation metrics for AI systems - Comfortable with the command line and version control workflows using Git (GitHub or GitLab), including branching, merge requests, and code review