Research Engineer, Character.ai

• As a passionate LLM data researcher, you love uncovering interesting subsets of data, creating clear dashboards to communicate your findings, and proposing ideas for unexplored opportunities
• You are also extremely interested in learning how to support large language model development with unparalleled data quality and performance insights
• You believe in data-centric AI, and know how to improve LLMs from the data perspective
• You want to be involved in training ever larger foundation models
• A bachelor's degree in a quantitative field (Physics, Mathematics, Computer Science); PhD preferred
• 2+ years experience in industry, training, and evaluating LLMs
• Experience mining text and graphical data
• Spark Experience

Desirable

• Data is the most important ingredient in training larger and better models. Getting the best data is critical to making our models more intelligent and empathetic. How good is our current data? What new data do we need to ensure our success? What data mixture should we use in the training runs?
• Curate, mine, and analyze datasets for LLMs
• Work with our pre-training/ post-training team to identify datasets needed for specific user experiences
• Run hundreds of careful ablation experiments to improve our models’ performance on benchmarks we care about
• Help maintain and improve core tables in our data lake used for research across the company
• Alongside the data platform team, you will be responsible for making sure this vital resource is available, understood, and of the highest quality