Roblox is looking to solve safety incidents and emerging trends at scale by leveraging innovative large-scale AI solutions to counter safety harms and risks across the Roblox platform.
Requirements
- 8+ years of experience with large-scale SRE, SecOps, MLOps, DataOps, or DevOps systems and operations in production environments.
- Proven track record of designing, developing, and launching rapid-response systems into production.
- 2+ years of hands-on experience using large language models or other AI/ML model technologies including prompt engineering & fine-tuning.
- Expertise in solving complex MLOps, data pipeline, and infrastructure challenges – with a focus on high velocity & quality at scale.
- Experience with modern microservice architectures and distributed systems paradigms (e.g. cloud services, scalable data pipelines, real time monitoring).
- Hands-on mindset: a willingness to dive into code/architecture reviews across the Roblox tech stacks, in addition to high-level planning.
Responsibilities
- Apply ML solutions to safety rapid response.
- Update our advanced ML models to detect emergent patterns of safety violations, leveraging repeatable training pipelines.
- Operate and enhance AI-powered safety tools leveraging multi-modal (text, image, video, behavioral, game context) detection strategies to detect and mitigate emergent safety violations, from real-time text filtering at scale to large language models for cutting-edge content moderation.
- Build and operate industry leading large-scale safety AI tools to understand complex user behaviors and implement interventions that mitigate abuse and harassment.
- Own the vision, direction, and execution of the safety rapid-response tools and technical operations for safety incidents, detecting and mitigating emergent patterns of violations at scale.
- Respond to emerging safety threats in hours & days, not weeks, by rapidly adapting and deploying the existing safety systems and ML models.
- Proactively engineer and improve rapid-response tools and systems
Other
- Lead and grow a high-performing team of rapid-response engineers, fostering a culture of Safety 24x7 vigilance and resilience, at the intersection of SRE, SecOps, MLOps, DataOps, and DevOps, for the Trust & Safety domain.
- In-take human insights (policy and new patterns) into Safety AI pipelines that can be rapidly scaled in production.
- Collaborate closely with product managers, policy experts, data scientists, and Trust & Safety operations, often wearing a product & operations hat yourself.
- Ability to thrive in ambiguity: you excel at bringing clarity and direction to undefined or open-ended problem spaces.
- Demonstrated success collaborating across functions (e.g. Product, Design, Data, Research, Ops), working together to drive meaningful business and user impact.
- Strong safety sense: able to establish clear success metrics and craft strategic roadmaps to achieve those goals.
- High emotional intelligence: adept at resolving conflicts, mentoring engineers, and nurturing the growth of your team members.
- Creative and strategic problem-solver: able to distill complex issues into simple, innovative solutions that drive impact.
- An owner: you take responsibility for projects and outcomes end-to-end, and instill the same accountability in your team.
- A starter & scaler: you are good at going from 0 -> 1 and exceptional at going from 1 -> 100.
- A lifelong learner: constantly seeking new knowledge and techniques to grow your expertise and expand your impact.
- Independent and self-directed: capable of charting a course with minimal guidance, and comfortable making decisions in uncertain situations.
- Highly collaborative: effective at working with cross-functional partners and teams, and skilled at building consensus.
- Committed to diversity & inclusion: dedicated to fostering a workplace where everyone feels valued, supported, and able to contribute.
- Passionate about cutting-edge AI / ML tech: knowledgeable about the latest advances in AI and excited to apply these technologies to keep users safe.
