Reasonable is the applied AI research company building formal verification for super-human software development.
Correctness guarantees for software developed by humans and machines are no longer impractical or prohibitively expensive. Code generated by AI can be provably correct, rather than plausibly functional. At Reasonable, we are doing the research, training the models, and developing the products required to make this a reality. Achieving this creates a new paradigm for high accountability software development and unlocks the full potential of AI for professional engineers.
We’re a compact, talent-dense technical team, with deep domain expertise in machine learning, formal verification and mathematical models of program semantics. Join us to develop the next frontier of formal reasoning and software engineering.
Proof follows function.
The Role
As a Member of Technical Staff, you will play a key early role at the core of Reasonable's research, engineering, and product development. Your work will shape the research vision and develop new capabilities at the frontier, where novel training approaches and formal methods intersect. Ultimately, your work will be instrumental in enabling formal oversight in software development.
Projects our team is working on include designing evals for state of the art coding models, developing novel post-training paradigms grounded in formal methods, and building the tooling to deliver correctness guarantees in production software engineering.
We’re an early-stage team tackling hard problems with varying degrees of predictability. Our roles require adaptability but, in return, we adapt to the candidate’s strengths. The entry point is depth in either machine learning or formal methods, alongside a strong software engineering background.
Requirements
We're looking for
- Domain expertise in either machine learning or formal methods, with active interest in learning the other
- Evidence of extremely fast learning of deeply technical subjects
- Experience running machine learning experiments, ideally at scale
- Experience post-training large language models
- Strong software engineering practice: advanced git workflows, testing, containerisation, code review, etc
- Familiarity with MLOps tools and training across multi-GPU clusters
- An understanding of specification-aware programming (Verus, Dafny, TLA+), proof assistants and verification tools ( LEAN, Isabelle)
- AI-natives, with experience using AI-assisted programming tools (Claude Code and similar)
Bonus Points If You
- Actively contribute to formal verification or program synthesis projects: Verus, Lean, Dafny, or similar
- Have run production back-end services at scale; you’ve felt the pain of what testing can’t prove
- Have been accountable for distributed systems; you respect the failure modes that emerge from concurrency, consensus, and partial failure
This is an unusual profile. If that’s you, get in touch. If you are close to it, we still want to hear from you! If you know someone that would be ideal, we always reward great introductions.
Benefits
- Compensation: Generous salary, with equity and additional benefits
- Location: San Francisco, with flexibility for team and conference travel - we're an on-site team
- Visa sponsorship: available for the perfect candidate
- World-class team and environment: an opportunity to build alongside deeply experienced founders, in a well-funded company that’s backed by industry leading VCs and angels including Oriol Vinyals, Zoubin Ghahramani, Jonathan Frankle and Guy Podjarny, amongst many others