Latest AI JobsBrowse Jobs
SASH logo

Research Engineer / Research Scientist, Remote Compute Accounting

SASHRemote/FlexibleVerification
Apply on company site
Salary
$100K – $200K
Job type
Full-time
Work mode
Remote
Posted
22 Sept 2026
Status
Open · checked 7d ago

About this role

Based remotely, this Research Engineer or Research Scientist position focuses on remote compute accounting for AI datacenters and suits individuals with backgrounds in GPU performance, systems measurement, or cryptography. SASH is hiring for this technical role to build and test verification tools that ensure AI clusters operate only on declared workloads.

Role Overview

This position sits within the verification team at SASH, focusing on software-only approaches to verify AI datacenter activity and prevent undeclared training runs. The role involves designing challenge protocols, measuring performance on real accelerators, and collaborating with cryptographers and hardware-security colleagues. The person in this role will own the problem of quantifying and minimizing unaccounted cluster capacity.

Key Responsibilities

Implement candidate challenge functions and characterize honest-response latency across actual cluster topologies.
Determine how much spare capacity a cluster can retain while passing challenges and minimize that margin.
Conduct adversarial compression studies on frontier model weights and KV caches.
Design and evaluate post-training methods that increase weight compression costs without harming model performance.
Quantify hidden memory capacity and develop protocol modifications to neutralize it.
Extend verification approaches from memory occupancy to compute using proof-of-useful-work schemes.
Execute protocols against live inference clusters and document failures.
Design red-team and blue-team competitions for the protocol without requiring new hardware.
Red-team internal timing assumptions.

Requirements

Experience in GPU performance or systems measurement, including kernels, microbenchmarks, memory hierarchies, and distributed timing, or a cryptographic background with a desire to work directly with hardware.
Rigorous approach to measurement and awareness of potential measurement vulnerabilities.
Ability to account for adversarial behavior during optimization.
Comfort working on experimental protocols that may fail and willingness to publish negative findings publicly.
Ability to thrive in ambiguous, early-stage environments where problem definition is required.
Willingness to maintain working-hour overlap with globally distributed colleagues across regions.

Nice to Have

Proficiency in CUDA or equivalent low-level accelerator programming.
Experience in compression, information theory, or model quantization.
Background in timing-based protocols, proofs of space or time, or distance bounding.
Experience characterizing undocumented hardware.
Familiarity with large language model inference internals such as batching, KV cache management, and scheduling.

Skills and Tools

CUDA
Low-level accelerator programming
GPU microbenchmarks
Memory hierarchies
Distributed timing
Cryptography
Information theory
Model quantization
LLM inference internals

Salary and Benefits

Salary range from $100K to $200K, with higher compensation available for exceptional candidates depending on location, experience, and level.
Competitive benefits and leave policies.
Visa sponsorship available in the UK for candidates interested in relocating.

About the Company

SASH builds and tests tools to verify agreements regarding AI. The organization prototypes verification mechanisms, leads international research collaborations, and works with global policymakers to demonstrate tool capabilities and inform usage. The technical team is led by Pascal Berrang, an Associate Professor at the University of Birmingham with experience from the Singapore Government, working alongside a policy team with backgrounds from Oxford and the Centre for the Governance of AI.

How to Apply

Candidates should apply directly on the official SASH careers page.

Skills and tech stack

CUDALLM

Questions candidates ask

What is the primary goal of the Research Engineer or Research Scientist role?

The role owns the challenge of determining how much of an AI cluster can be accounted for from the outside and minimizing any leftover room to prevent undeclared training runs.

Where is the position located?

This is a remote role that can be based in Singapore, the UK, or remotely, with requirements to maintain working-hour overlap with globally distributed colleagues.

Is visa sponsorship available?

Yes, SASH provides visa sponsorship in the UK for candidates interested in relocating.

What is the salary range for this position?

The stated salary range is between $100K and $200K, with higher amounts available for exceptional candidates depending on experience, level, and location.

What kind of professional background is expected?

Candidates are expected to have GPU performance or systems measurement experience, or a cryptographic background with an appetite for working directly with hardware.

Ready to apply?

The application is on the official SASH careers page.

Apply on company site