Research Engineer

ABOUT RESOLUTION Resolution does research on how to align artificial superintelligence (ASI). ASI may be developed in the next few years, but it is unclear whether alignment is on track to be ready in the same timeframe. We aim at higher a priori confidence in aligned outcomes by pursuing a portfolio of theory and empirics bets, any one of which — if it succeeds — would meaningfully advance the field. We invest heavily in research automation to accelerate progress, and we believe that stronger alignment theory unlocks higher automation: more principled approaches give us better filters for which directions of automated research are promising. Resolution was founded in 2026 by researchers from UK AISI's Alignment Team, who ran the £30m Alignment Project https://alignmentproject.aisi.gov.uk/, and Timaeus https://timaeus.co/, who pioneered applying singular learning theory to alignment. For more information, see our announcement https://sequent.org/launch. ABOUT THE TEAM We are hiring Research Engineers across several main focus areas. We expect the boundaries between these areas to be flexible, but please indicate which mode you're more interested in (or "either") in your application. Research Automation (primary focus). A cross-cutting function that builds the infrastructure and tooling our researchers use to scale their work, increasingly leveraging fleets of AI research assistants alongside small teams of humans. Program-embedded Research Engineering (also hiring). Research engineers embedded within one of our research programs (scalable oversight, complexity theory, learning theory, personas, and possible future programs like heuristic arguments or game theory), partnering with researchers on scaling experiments, building program-specific infrastructure, and translating theoretical insights into empirical tools. ABOUT THE ROLE Research Engineers at Resolution are core members of our research teams, directly driving both research and the core infrastructure behind it. We believe clean engineering on automation, experimentation, and infra is essential to ambitious research, and that excellence on this front requires active research participation. RESPONSIBILITIES - (Research automation track) Build agentic research infrastructure: experiment orchestration, hypothesis generation, automated analysis pipelines; autoformalization tooling for the theory side; internal AI-powered tools for researchers. - (Program-embedded track) Scale program experiments to frontier-tier models; build program-specific infrastructure; partner with researchers on engineering and implementation. - (Both) Maintain and extend distributed training, experiment, and evaluation infrastructure. - (Both) Contribute to and maintain shared codebases across the org. - (Both) Communication of engineering & automation progress, obstacles & learnings to your team and the wider org via Slack and in weekly meetings. We're on the lookout for excellence, so if you're a cracked engineer who doesn't precisely fit these descriptions, please still apply! YOU MAY BE A GOOD FIT IF YOU - Have a strong software engineering background, including production-quality Python - Have deep experience with ML frameworks (PyTorch or Jax) and distributed-training stacks - Have a demonstrated ability to ship complex systems end-to-end - Have a Bachelor's degree or equivalent in CS, physics, math, ML, or related - Are willing to use AI tools aggressively in your own workflow, with appropriate care to not get fooled! - Are motivated by alignment of artificial superintelligence (ASI) and want to contribute to it full-time. STRONG CANDIDATES MAY ALSO HAVE - Experience with autoformalization, Lean, or other proof-assistant tooling - Background in research infrastructure or ML platform engineering at frontier labs - Experience scaling ML systems to 100B+ parameter scale - Experience with CUDA kernel development or GPU optimization - Familiarity with alignment research APPLICATION PROCESS 1. Application review. Every application gets at least one human review. 2. CodeSignal (90 minutes). The Industry Coding Assessment, not LeetCode: we want the content of our assessments to have at least some overlap with the things we expect you to do in real life, even if asking you to write code is a bit dated. 3. Screening call (15 minutes). An informal chat, where we briefly cover background, fit, and motivation. 4. Technical interview #1 (45 minutes). This will likely be a performance-optimization-style problem. We'll give you some demo code and ask you to spot issues and propose solutions, without AI assistance. 5. Experience interview (60 minutes). We'll walk through your background in much more detail, asking you to go through one or more past projects you've worked on. We'll also take more time to discuss the role and answer your questions, and we'll share more information before you get to this stage. 6. Technical interview #2 (60 minutes). This has two parts: (a) an AI-unassisted systems design and specification problem, followed by (b) an AI-assisted implementation of (a). 7. 1-on-1s. A chat with our chief scientist, Geoffrey, as well as any relevant people from the team you would like to talk to and ask questions. 8. References. We'll reach out to a few references as we make our final decision. On scheduling. In some cases, we'll bundle (3) and (4) so they occur back to back. In almost all cases, we'll bundle (5) and (6), though with different interviewers, so you get a chance to talk to more of our team. We'll also run (7) and (8) in parallel. Our aim is to get you an update within 2-3 days of each stage, which means a process of about three weeks. For some candidates, especially if you have competing deadlines, we'll aim to push faster. On the technical interviews. We dropped work tests because there's so little signal left in them. What we're looking for is how well you figure out what to tell the AIs (6a), how well you interact with the AIs throughout impl