Research Engineer/Scientist, Remote Compute Accounting

Design and test verification protocols to detect undeclared AI model training in datacenters using hardware measurement and cryptographic approaches.
- Microbenchmark GPUs and design challenge protocols to measure cluster capacity and detect spare hardware resources.
- Characterise memory bandwidth, compression feasibility, and timing under adversarial conditions to close loopholes in verification schemes.
- Collaborate with cryptographers and hardware-security experts to construct hardness assumptions and publish findings, including negative results.
- Strong GPU performance engineering, systems measurement, or cryptographic background with appetite for hands-on hardware work.
- Comfortable in ambiguous early-stage environments; rigorous about measurement and adversarial reasoning.

SASH is a Singapore-based nonprofit that builds AI safety research capacity and governance expertise across Asia.