Why Safety?

At Roblox, we strive to connect a billion people with optimism and civility, and the Safety organization’s mission is to become the leader in civil immersive online communities. We systematically and proactively work to detect, remove, and prevent problematic content and behavior. We seek to influence and shape the product roadmap and prioritization, build safety products, and measure our impact on the community of users and developers. In doing so, we help keep Roblox safe, civil, and inclusive, fostering positive relationships between people around the world.

Why Safety Core Data?

Safety data is one of the hardest—and most impactful—frontiers in machine learning, and in this role you’ll advance large-scale AI systems that safeguard platform integrity and user trust by building our Safety Foundation Model. You’ll leverage state-of-the-art multimodal models to create intelligent tooling for curating training and evaluation data, enabling continuous improvement and proactive, real-time mitigation of safety risks. Operating at the intersection of foundation models, agents, synthetic data, and platform safety, you’ll help design the core infrastructure that allows models to learn from real-world errors through a ship-often, feedback-driven MLOps discipline.

As a senior machine learning engineer, you’ll work hands-on with cutting-edge systems alongside experienced ML and Infra engineers, product managers, policy and safety experts to define technical standards and best practices for the next generation of safety AI, translating complex safety challenges into scalable, high-impact machine learning solutions. Your contributions will directly power production-grade moderation, detection, and enforcement systems, helping keep the platform secure, inclusive, and resilient.

You Will

• Design, build, and own core components of our Safety AI Systems, powering large-scale safety data generation, labeling, and evaluation with a focus on reliability, scalability, and performance.


• Fine-tune open-weight multimodal LLMs (text, image, video, and voice) and deploy them as agents to generate synthetic datasets and perform high-quality automated labeling at scale.


• Build robust pipelines for curating and evaluating multimodal training data, including human-in-the-loop systems that enable safe, compliant, and continuously improving AI.


• Develop benchmarks and evaluation frameworks to measure multimodal model behavior, data quality, and downstream safety impact.


• Architect systems that support continuous learning, enabling weekly gated rollouts and tight feedback loops from real-world failures.


• Partner closely with engineers, product managers, policy experts, and safety specialists to translate complex safety challenges into scalable, production-grade AI solutions.


• Help define best practices, technical standards, and data culture for next-generation safety AI, building foundational systems from the ground up and driving impact quickly.

You Have

• 5+ years of experience designing, building, and deploying large-scale machine learning systems in production environments.


• A graduate degree or equivalent experience in Computer Science, Engineering, or a related technical field.


• Strong communication skills and a collaborative, solution-oriented approach to problem solving.


• Hands-on experience with fine-tuning VLMs, LLMs, or large multimodal models to improve model quality, safety, and performance is a plus.


• Experience working with data quality, evaluation, synthetic data, or ML infrastructure in safety systems is a plus.

You Are

• Data & System Oriented: You understand that robust data and systems are the foundation of any production applications, and you design infrastructure for scale, correctness, and reliability.


• Collaborative: You thrive in cross-functional teams, partnering effectively with Product and Engineering to solve complex challenges.


• Curious & Creative: You enjoy tackling hard problems, exploring new technologies, and driving continuous improvements in both systems and workflows.


• Impact Driven: You focus on outcomes - prioritizing product impact, reliability, and measurable success over code volume or complexity.