SlipstreamJobs tracks this role from the company's public career site. Apply directly on the employer's site.
Salary: USD 350,000 - 850,000 / annual
Anthropic is seeking a Research Engineer to lead the end-to-end development of training data and reinforcement learning environments for visual knowledge work capabilities in large language models. This role sits at the intersection of applied research and hands-on data engineering, focusing on multimodal AI systems.
Key responsibilities include owning the data strategy for vision capabilities from conception through scaling, managing technical relationships with external data vendors to ensure high-quality visual annotations and reward design, developing QA frameworks to detect reward hacking and maintain environment quality at scale, and running generalization experiments to validate how data strategy improvements translate to real-world performance gains.
You will collaborate closely with pretraining, reinforcement learning, and product teams to ensure that the environments and datasets you build directly improve Claude's ability to perform complex visual reasoning tasks. The role requires identifying long-horizon, vision-heavy tasks suitable for RL training, designing robust evaluation metrics, and iterating rapidly on reward mechanisms based on empirical results.
Ideal candidates bring 7+ years of experience in machine learning, computer vision, and software engineering, with demonstrated expertise in reinforcement learning, reward design, or training data curation for large language or vision-language models. Strong familiarity with the architecture and training of large vision-language models is essential. You should be comfortable managing vendor relationships, writing detailed technical specifications, and evaluating annotation quality at scale. Experience designing benchmarks for LLMs, large-scale pretraining/RL, deep learning on images/video, agentic systems, or ETL pipelines is highly valued.
The role offers significant autonomy and impact, requiring a results-oriented mindset with flexibility to adapt as research priorities evolve. Anthropic values candidates who care deeply about the societal implications of AI systems.
