Founding Researcher
SAN FRANCISCO (IN PERSON)
bylaw’s mission is to ensure AI follows the law.
ABOUT BYLAW
Over generations, humans have crafted laws that, while imperfect, translate high-level values into evolving rules. These rules allow multi-actor systems, such as financial markets, civil societies, and democratic institutions, to function and flourish.
We need AI to follow these rules, both as a baseline for AI safety and as a tool for advancing it. AI must follow the law to protect these foundational systems; legal alignment also helps prevent other safety concerns, such as deception and collusion, because documented past bad human behavior provides a measurable template to guide complex reasoning.
From AI safety policy in the Biden Administration to alignment research at Scale AI, we have witnessed firsthand a gap in the AI safety ecosystem: alignment to the laws that we, as a society, have crafted together through democratic processes.
ABOUT THE ROLE
bylaw is seeking a researcher to advance our research on law-following AI. You’ll design and run experiments across our three priorities: testing agents against the law, training them to follow it, and controlling them at runtime. As one of our first researchers, you’ll own projects from conception to publication, working closely with our founders, research lead, and academic advisory board. This is a full-time, in-person role, based in San Francisco.
QUESTIONS YOU’LL WORK ON (TO NAME A FEW)
- How do we turn court decisions, and the facts and reasoning they record, into realistic simulated environments, and generate new scenarios automatically from the rules they state?
- How should agents reason across conflicting obligations, overlapping jurisdictions, and open-textured standards that leave room for discretion, such as "reasonable," "material," or "clear and conspicuous"?
- How do we validate evaluations of legal compliance when the judge is itself a model, and how do we build labeled data from court decisions to train better judges?
RESPONSIBILITIES
- Design and run experiments across bylaw’s research agenda.
- Build evaluation environments, training pipelines, and runtime harnesses.
- Work with lawyers and our advisory board to ground research in real legal practice.
- Publish research and release benchmarks and tools.
WHAT YOU WILL DO IN YOUR FIRST YEAR
- Lead a research project from design to publication.
- Build legal evaluation environments and benchmarks grounded in real court decisions.
- Develop labeled data and methods for training and validating LLM judges.
REQUIREMENTS
- Degree in computer science, machine learning, or a related field, or equivalent practical experience.
- One or more years of research or ML engineering experience, in industry, academia, or independent projects.
- Experience in AI safety, alignment, or evaluations, through publications, open-source projects, or industry work.
- Strong engineering skills in Python and the modern ML stack.
- Authorized to work in the US now and in the future without visa sponsorship; we are not able to sponsor visas at this time.
WHO YOU ARE
- A deep commitment to AI safety and to making AI follow the law.
- An excitement for tackling challenges with no predefined solution.
- A drive to move at startup pace without sacrificing rigor.
NICE TO HAVE
- Legal training, or research in law and computation.
- Experience training or validating LLM judges.
- Hands-on experience building evaluations, training environments, or post-training pipelines.
JOIN US
We are a group of lawyers, alignment researchers, and policymakers from Stanford, UC Berkeley, and Scale AI working to align AI to the law. We seek top researchers to join us in our mission. Contact christina@bylaw.com if you are interested in learning more.