Principal Data Scientist - Safety

Written by the T&S Network from Roblox's posting.

Quantitative safety work for user-generated content on Roblox, covering games, avatars, and emerging in-game behaviors. The team uses data science, statistics, and causal inference to define, measure, and improve detection and mitigation of harmful content and violative on-platform behavior, with a focus on establishing ground truth datasets and prevalence measurement.

What you'd do

• Establish and monitor prevalence measurement methods to quantify harmful content and behavior across the platform.
• Run exploratory analysis of current and emerging threats and recommend where Product and Engineering should invest detection and moderation resources.
• Design and analyze experiments for new moderation and enforcement features.
• Apply causal inference to measure effectiveness and systemic impacts of safety initiatives on player experience and integrity.
• Brief leadership and cross-functional partners on prevalence, reference data reliability, and how well safety measures work.
• Partner with ML and Data Engineering teams so detection systems rest on sound ground truth and measurement frameworks.

What it asks for

• 7+ years using Python or R and big data tools such as SQL, Hive, Spark, and Airflow.
• Knowledge of machine learning and deep learning through training or industry experience.
• Experience building large-scale safety or moderation systems.
• Experience with user-generated content platforms.

Terms in this role

Prevalence.

How often people see violating content, usually counted as views of it per 10,000 views. A common way to measure enforcement.

Our best summary of the posting, and it can be wrong. The employer's posting is what counts: read it in full on Robloxnorth_east before you apply. If we got something wrong, report this listing.