How to apply for Research Engineer - AI Verification

SASH

About SASH

SASH is a Singapore-based nonprofit building AI safety research capacity and governance expertise across Asia. It runs the first international collaboration on AI verification, with partners including the Future of Life Institute and University of Oxford. The team has presented at ICML, the AI Security Forum, and the World AI Conference.

About the role

This role is on SASH's Verification team, building public prototypes of tools that verify what happens inside datacenters running large AI models. The work feeds into policy change: the team demonstrates tools to policymakers, roadmaps production-scale verification, and grows the base of independent experts who can evaluate these tools.

A typical day

A typical day likely involves building or testing a verification prototype, writing public documentation or a demo, and discussing with international partners or policymakers. Because the role is remote and the team is spread across institutions, much of the work happens in async writing and calls. Ask the team directly about their daily routines and meeting cadence.

Who SASH is looking for

  • Has hands-on experience building research prototypes or systems for AI verification, model internals, or hardware/datacenter monitoring.
  • Can work in public: writing code, docs, and demos that outside experts and policymakers can inspect and evaluate.
  • Comfortable with international collaboration across time zones and with partners from institutions like Oxford and the Future of Life Institute.
  • Interest in the policy side of AI safety: international agreements, middle-power sovereignty, and adoption in high-stakes industries.

Tips for this application

  • Read SASH's published verification work and reference at least one specific prototype or presentation in your application.
  • Describe one project where you built something in public (open repo, paper, demo) and what others could verify from it.
  • Name the verification methods you have used, such as model internals, hardware telemetry, or datacenter monitoring, and what you would want to learn next.
  • Explain how your work would be usable by policymakers or independent evaluators, not just researchers.
  • Since the role is remote and international, state your working hours and how you have collaborated across time zones before.

What to cover in your cover letter

A cover letter should cover: (1) a concrete verification or AI systems project you built and what it showed; (2) why public prototypes and policy translation matter to you; (3) how you have worked with international or cross-institutional partners; (4) what you want to build at SASH and which part of the Verification team's work you would contribute to.

Draft a cover letter

Research before applying

  • Read SASH's public writing and the response mentioned in the job description about their summer presentations.
  • Look at the Verification team's prototypes and any public repos or demos.
  • Review the work of partner organizations: Future of Life Institute, University of Oxford, Centre for the Governance of AI.
  • Check the AI Security Forum, ICML, Australia AI Safety Forum, and World AI Conference talks SASH participated in.
SASH website

Likely interview topics

Based on the job description, expect questions about:

  • How you would design a prototype that verifies what is happening inside a datacenter running large AI models.
  • Your view on what makes a verification mechanism trustworthy across countries, not just within one.
  • A past project where you built in public and how external experts evaluated it.
  • How you would explain a technical verification tool to a policymaker.
  • How you would work with partners such as the Future of Life Institute, University of Oxford, Arm, or Intel.
Practise interview questions

Common mistakes to avoid

  • Treating this as a standard ML engineering role and ignoring the verification and policy-translation focus.
  • Claiming interest in AI safety without naming any specific verification method, tool, or paper.
  • Applying without checking whether you can work with a Singapore-based nonprofit and its international partners across time zones.

Deadline

No deadline is listed. Roles without a deadline usually close once the employer has enough candidates, so apply soon if you are interested.