The Role
As a member of technical staff, you will lead the core engineering effort to build compute efficient, physically accurate world models. Your work will enable our AIs to "dream" and predict how the world evolves, and help engineers evaluate and debug physical AI systems in real-like environments.
You own the inference lifecycle of the World Model, ensuring that execution performance is never the bottleneck for research iteration or production-grade deployment. You have a deep understanding of the high-performance runtime stack, and you are capable of delivering the low-latency, controllable sequence generation required for long-horizon rollouts across distributed clusters and target environments.
Job Responsibilities
- Area: Design and scale high-performance inference systems for large-scale generative world models.
- Ownership: The delivery of real-time predictive performance and execution reliability within physical control loops.
- Focus: Optimizing complex generative architectures to meet the stringent latency and throughput requirements of high-frequency systems.
- Operations: Establish the benchmarking and profiling standards that guide the architectural trade-offs between model complexity and runtime efficiency.
Minimum Qualifications
- Extensive experience in software engineering for performance-critical deep learning R&D teams.
- Expert-level experience with high-performance inference runtimes and the optimization of large-scale generative models for low-latency environments.
- Comfortable with Python, C++ / Rust.
Preferred Qualifications
- Proof of outstanding contribution to open-source libraries for high-performance inference, model serving, or deep learning compilers.
- Peer-reviewed research contribution in efficient machine learning (e.g., quantization, distillation, or efficient generative modeling) from a AAA conference.
- Experience with low-level hardware optimization and kernel development.