Member of Technical Staff

Member of Technical Staff, Research — Multimodal Models & Video Understanding We are partnering with a stealth AI research lab on a foundational research hire. Founding team includes leaders and founding members of Meta Superintelligence Labs, Tesla AI, and Physical Intelligence. About the company Their mission is to expand what humans and AI can accomplish together. We're entering an economy where value increasingly comes not from what we produce alone, but from the combined output of people and the AI systems working alongside them. Realizing that potential requires AI to understand far more than text, it must understand the world in which people actually operate: the decisions they make, the context they use, and the knowledge embedded in how things get done. Most of that knowledge has never been written down. It's tacit, distributed across people, screens, tools, and the physical world. Today's AI systems understand only a narrow record of human knowledge, rather than the full reality from which that knowledge comes. This company exists to close that gap. Its core technology is a new class of multimodal models that can capture, understand, and codify knowledge that doesn't otherwise exist in a form machines can use. The first application is work: building agents that learn how complex work is performed and help people carry it out. The role We're looking for an exceptional research engineer or research scientist to push the frontier of multimodal intelligence and video understanding. You’ll develop models that can understand long-horizon activity, reason across time, extract structure from complex environments, and convert multimodal experience into representations that language models and agents can use. This is a foundational research role. You’ll help define the architectures, training methods, datasets, and evaluations required for AI systems to move beyond the written record and begin understanding the world itself. What you’ll work on Develop multimodal foundation models that learn from video, language, audio, and interaction data, with a particular focus on long-horizon understanding and temporal reasoning. Advance representation learning, context compression, multimodal alignment, and architectures that transform complex activity into structured knowledge and action. Own ambitious research directions end to end — from initial hypothesis and data strategy through large-scale training, rigorous evaluation, and integration into production systems. What we’re looking for Exceptional research ability in multimodal learning, video understanding, computer vision, language modeling, or a closely related field, with experience training modern models at meaningful scale. A record of driving an important project at a premier company or academic institution, with clear ownership of its direction and outcome. Strong judgment across architecture, data, experimentation, and evaluation, combined with the independence to operate where the correct approach is not yet known. Strong candidates may come from organizations such as Luma AI, Runway, Pika, Google DeepMind, OpenAI, Anthropic, xAI, or an exceptional academic research institution. Publications are valuable but not required — what matters most is evidence that you can identify an important problem, develop a rigorous approach, and drive it to a meaningful result. You’ll work on one of the central unsolved problems in AI: building models that understand the world beyond the fraction of it captured in text. Your research won’t remain isolated. It will become the intelligence behind products operating in real environments, creating a direct path from foundational research to systems that expand human capability. #J-18808-Ljbffr