Roblox is looking to solve high-scale language-related problems on its platform, including real-time moderation of voice and text, automatic localization of experiences, and empowering users with LLM-driven creation tools, by building and deploying cutting-edge NLP, speech, and generative AI models.
Requirements
- Expertise in one or more areas: NLP, Speech Models, Large Language Models, Machine Translation, or Generative AI (including diffusion models).
- Experience with transformer-based model design, training, serving, and product integration.
- Proficiency in one or more programming languages (e.g., Python, C++, Go, Java) and experience building and optimizing large-scale systems.
Responsibilities
- Design and implement deep learning-based NLP and speech solutions that address problems across Roblox, from creation to safety.
- Develop advanced models, including Large Language Models (LLMs), machine translation, and generative AI, for user interactions, content creation, and moderation.
- Have the independence and end-to-end responsibility to develop NLP/ML-based services that are scalable and resilient.
- Be a technical bar-raiser for cutting-edge ML technology, high code quality, and architectural designs.
- Work backward from user and product needs to deliver ML solutions that drive engagement, safety, and ecosystem growth.
- Building world-class NLP and speech models for real-time moderation of voice and text (processing over 6 billion messages daily) and advanced interventions that measurably improve user civility.
- Developing specialized LLM agents, behavioral analysis, and graph systems for detecting and preventing rare, high-risk scenarios (e.g., child safety, terrorism), requiring adversarial thinking and multi-step reasoning.
Other
- Possessing or pursuing a Ph.D. in Computer Science, Computer Engineering, Mathematics, Statistics, or a related technical field, with a thesis aligned to Roblox’s research areas.
- A strong research track record, evidenced by multiple publications and presentations in top-tier, peer-reviewed venues (e.g., ACL, EMNLP, Interspeech, ICML, NeurIPS).
- Roles that are based in an office are onsite Tuesday, Wednesday, and Thursday, with optional presence on Monday and Friday (unless otherwise noted).
