Member of Technical Staff, Forward Deployed Engineer

Required

• We’re hiring Forward Deployed Engineers to bring Odyssey’s world models into high-impact, real-world environments.
• This role sits at the intersection of research, engineering, product, and our partners. You’ll work directly with customers to understand problems that cannot be solved with conventional software, translate those problems into research questions, and build working systems around Odyssey’s models.
• This is not a traditional research engineering role. You’ll be expected to modify models, design experiments, build new conditioning and control methods, develop evaluation frameworks, and turn ambiguous partner needs into technical breakthroughs. The strongest work will both solve an immediate application and influence Odyssey’s core research roadmap.
• Deeply technical, with experience in world models, generative video, multimodal learning, robotics, autonomous systems, reinforcement learning, computer vision, or large-scale machine learning.
• An unusually strong builder. You can move from an ambiguous problem to an experiment, prototype, and deployed system without waiting for a detailed specification.
• Comfortable modifying models as well as building the systems around them. You can work across training, post-training, inference, data pipelines, evaluation, and application code.
• Skilled at working from first principles. You can distinguish a genuine research problem from an integration problem and design the right intervention for each.
• Fluent across the research-to-production gap. You care why a model works, but also whether it works reliably on real data, at acceptable latency, within operational constraints.
• Effective with customers and external partners. You listen carefully, develop a precise understanding of their needs, and establish credibility through technical depth and execution.
• Able to manage competing priorities without losing sight of the underlying research opportunity or customer outcome.
• Energized by unfamiliar domains. You can quickly learn the relevant workflows, terminology, constraints, and failure modes.
• Comfortable operating in high-trust, high-autonomy environments where the path is unclear and iteration speed matters.