Research Engineer, AI Verification

The Team

The world is waking up to the fact that we will need ways to verify what's happening inside datacenters running large AI models -- to enable international agreements, protect middle-power sovereignty, and facilitate trustworthy adoption of AI in high-stakes industries. But, for this to be trusted, it can't just be developed in a few countries.

Singapore AI Safety Hub is launching the first international collaboration aimed at changing that. Our Verification team builds prototypes of these tools in public to speed up the development and adoption of globally trusted verification mechanisms. Our current effort spans Singapore, USA, UK, Canada, and Germany, but we're looking to expand even further.

We're building tools that will translate into policy change in the real world because our team is doing more than just building. Our team is demonstrating these tools to policymakers globally, helping roadmap the path to production-scale verification mechanisms, and broadening the base of independent experts who can evaluate the these tools.

Our partners include experts from the Future of Life Institute, University of Oxford, and more. Our core team has experience at Oxford, ByteDance, Centre for the Governance of AI, and Singapore Government. Our collaborators have worked with Arm, Intel, and RAND.

Just this summer, we plan to present our work at the AI Security Forum (Washington D.C.), ICML (Seoul), Australia AI Safety Forum (Sydney), and World AI Conference (Shanghai)

Your Work

You'd play a key role in building prototypes and supporting technical communications for our growing Verification team. Developing strong in-house capacity will be essential for steering this collaboration and informing what gets built.

Work spans the gamut — from developing algorithms for recomputing AI workloads, to ML engineering to set up inference pipelines, to frontend engineering of demos, to cybersecurity analysis of our designs. In all cases, you'd get to collaborate with external experts and the rest of our team.

• Our current project involves distinguishing between inference and training workloads on a GPU.
• Potential future projects:prototyping secure GPU enclosures, or designing privacy-preserving means of conducting white-box evaluations.

About You

Essentials

• Strong ML engineering fundamentals — we care more about ability and trajectory than years of experience.
• Familiarity with key concepts in AI verification, cybersecurity, and/or confidential computing.
• Thriving in ambiguous, early-stage environments where defining the problem is part of the job.

Nice to have

• Experience contributing to AI hardware engineering or confidential computing projects.
• Cybersecurity experience, particularly around network-level instrumentation.
• Experience as technical staff at an early-stage organisation.
• Familiarity with, or interest in working in, Singapore.