[Hiring] Research Engineer / Research Scientist @Singapore AI Safety Hub

Role Description

The team is focused on answering a critical question: can we verify what is happening inside AI datacenters? This could unlock international cooperation, help countries protect their sovereignty, and enable trustworthy adoption of AI in high-stakes industries.

SASH’s Verification team builds and tests tools to verify agreements about AI. Key activities include:

• Prototyping new verification mechanisms
• Leading international research collaborations
• Working with policymakers globally to demonstrate tool capabilities and inform usage

You would join a small team combining technical expertise with international AI policy experience, led by:

• Pascal Berrang, Associate Professor at University of Birmingham
• Team members with experience from the Singapore Government, Oxford, and the Centre for the Governance of AI
• Partners including experts from the Future of Life Institute and the University of Oxford

Requirements

• Designing challenge protocols and measuring whether real accelerators can satisfy them
• Microbenchmarking and writing challenge kernels
• Characterising high-bandwidth memory under adversarial conditions
• Deciding whether the numbers support a claim for an agreement
• Working alongside cryptographers and hardware-security colleagues
• Publishing findings, including negative results

Representative Projects

• Implement a candidate challenge function and characterise honest-response latency across real cluster topologies
• Establish how much spare capacity a cluster could retain while still passing every challenge
• Run the adversarial compression study on frontier model weights or KV cache
• Design and evaluate post-training to raise the cost of compressing weights
• Quantify how much memory a prover could plausibly hide
• Extend the approach from memory occupancy to compute through proof-of-useful-work schemes
• Take the protocol end to end against a live inference cluster
• Design a red-team/blue-team competition for the protocol
• Red-team timing assumptions

About You

We are hiring this role at a range of levels, prioritizing ability and trajectory over years of experience. You may be a good fit if you:

• Have GPU performance or systems measurement experience or a cryptographic background
• Are rigorous about measurement and aware of potential pitfalls
• Can conceptualize adversarial scenarios while optimizing
• Are comfortable working on potentially non-functional protocols and discussing them publicly
• Thrive in ambiguous, early-stage environments where defining the problem is part of the job

Strong candidates may also have:

• CUDA or equivalent low-level accelerator programming experience
• Knowledge in compression, information theory, or model quantisation
• Experience with timing-based protocols, proofs of space or time, or distance bounding
• Experience characterising undocumented hardware
• Familiarity with LLM inference internals, including batching and KV cache management