What you will be working on
The AI Enablement Office (AEO) aims to enable and drive AI adoption across the Board. The AI Performance Team, part of AEO, works hand-in-hand with product teams across CPF Board (CPFB) to evaluate and improve the performance of AI products — from defining what "good" looks like, to diagnosing where AI systems fall short, to closing the gap so they work reliably in production. Alongside this, the team builds organisational capability in AI performance engineering through the development of guidelines and playbooks.
You will be part of early-stage, exploratory work, helping to make sense of a problem space and figuring out together what, if anything, should be built. This is not a role with a neatly scoped project waiting for you. If you thrive in ambiguity, enjoy questioning assumptions, and find the messy front-end of problem-solving energising rather than frustrating, this might be the right fit.
In this role, you will:
- Work with product teams across the AI software development lifecycle (SDLC), from scoping to post-deployment.
- Build evaluation datasets, including golden sets and edge cases from real usage.
- Set up evaluation pipelines using LLM-as-judge, human review, and automated tests.
- Set up post-deployment monitoring to catch drift and regressions early.
- Diagnose failure modes (wrong tool selection, missed context, inconsistent outputs) and fix prompts, tool definitions, and agent workflows.
- Define quality standards and release criteria with product and engineering teams.
- Develop playbooks and templates so teams can run evaluations on their own.
What we are looking for
We value the diverse skills and perspectives that each intern brings. While you may not need to meet every requirement fully, having some familiarity or budding expertise in the following areas will help you make the most of this opportunity and succeed with our team.
- Currently pursuing a Degree in Computer Science, Information Systems, Business Analytics, Data Science, or a related field.
- Familiarity with Large Language Model (LLM) APIs and core concepts (tokens, context windows, temperature, tool use).
- Hands-on experience with frameworks like LangChain and LangGraph is advantageous but not essential.
- Some experience or genuine interest in coding, low-code tools, APIs, or AI-assisted development; familiarity with JavaScript/TypeScript, Python, or similar is a bonus, not a requirement.
- Analytical mindset with an interest in user research and testing; able to frame simple experiments, make sense of information, and draw tentative conclusions, including designing surveys or feedback mechanisms, analysing user behaviour and adoption patterns, and providing actionable insights to improve digital products.
- Curious about AI and LLMs, with an interest in how they work, where they fall short, and how to think critically about their outputs and real-world usefulness.
- Comfortable with ambiguity and honest about it; able to make progress without a fully defined brief, without overstating clarity you don't have.
- Inherently curious, thoughtful, and willing to challenge assumptions and communicate views openly, including knowing when to pause and check in.
- Works well independently and with others, takes initiative without needing constant supervision.
Position is on a full-time internship basis from January 2027 to June 2027.