Member of Technical Staff - UK Talent Pool

Reasonable is an applied research company developing training paradigms for superhuman programming AI. We draw on on domain expertise in machine learning, formal verification and mathematical models of program semantics to create open-ended training environments, pre-training datasets, and post-training methods that continue to challenge LLMs even as they develop deeply superhuman levels of understanding and skill in programming.

We think when AI writes most code, it will ultimately enable new ways for computers to be programmed, using programming paradigms that are just too challenging for humans to use at scale, enabling optimisations that were previously too complex to attempt. We call these applications post-human software engineering.

We aim to create a compact, talent-dense technical team to develop the next generation of frontier training methods.
We want to build in Europe, have impact in the Bay Area. Currently we prioritise applicants interested in relocating to Budapest.
We plan to scout for talent in the London, UK and San Francisco, US areas in the near future — if you're interested, please sign up.

The ideal candidate demonstrates:

• domain expertise in either machine learning or formal methods; interest in learning the other if expertise is in one of these
• fast learning of deep technical subjects
• experience running machine learning experiments, ideally at scale
• experience with post-training large language models
• knowledge of software engineering best practices (advanced git workflows, testing, containerization, code reviews, etc)
• familiarity with MLOps tools, working models on multi-GPU clusters
• an understanding of specification-aware programming (Dafny, Viper) and proof assistants (LEAN, Isabel)
• experience using AI-assisted or AI-accelerated programming (Cursor or similar)

and/or a willingness and ability to learn and grow in any of the above areas and beyond.

• Competitive salary
• Equity (through share option scheme)
• Flexible working patterns
• Healthcare, childcare, fitness and other common benefits (we're figuring this out)
• Being there from the very early stages