TL;DR
Member Of Technical Staff (AI): Developing open-weight models and general agents with an accent on post-training optimization, data curation, and reinforcement learning. Focus on building data generation pipelines, reward models, and inference-time scaling techniques to improve reasoning and instruction-following capabilities.
Location: Must work on-site in San Francisco, New York, or London
Company
An AI startup building open superintelligence with a team of researchers from leading industry organizations.
What you will do
- Build systems that transform pre-trained models into aligned, general-purpose agents.
- Drive research initiatives spanning from data curation to large-scale optimization.
- Develop and refine data generation pipelines, reward models, and RL algorithms.
- Collaborate across pre-training and post-training teams to achieve step-function gains in capability.
- Analyze and shape the understanding of how large models learn to reason and follow instructions.
Requirements
- Deep understanding of machine learning fundamentals and practical experience with large-scale LLM training.
- Strong engineering skills with experience in complex ML codebases and distributed systems.
- Proven ability to improve model behavior through data, reward modeling, or RL techniques.
- Evidence of leading research or engineering agendas that resulted in measurable model improvements.
- Ability to work fluidly across research and infrastructure boundaries.
- Strong communication skills and experience working in collaborative, high-agency environments.
Culture & Benefits
- Top-tier salary and equity packages designed for industry-leading talent.
- Comprehensive health, dental, vision, life, and disability insurance.
- Fully paid parental leave and financial support for family planning.
- Relocation support for eligible candidates.
- Daily catered lunch and dinner with regular team off-sites and celebrations.
Будьте осторожны: если работодатель просит войти в их систему, используя iCloud/Google, прислать код/пароль, запустить код/ПО, не делайте этого - это мошенники. Обязательно жмите "Пожаловаться" или пишите в поддержку. Подробнее в гайде →