Infrastructure Numerics Research Engineer for Large-Scale AI

Bison; San Francisco (CA); Presencial; USD 400.000 - 500.000; Jornada completa

Ventajas ofrecidas por este puesto de trabajo

Health insurance

Dental

Vision

Unlimited PTO

Parental leave

Relocation support

Descripción de la vacante

Thinking Machines is hiring an infrastructure research engineer in San Francisco to design and build core systems for efficient large-scale model training. You will optimize distributed training stacks, implement low-precision numerics, and develop kernels with hardware-enabled communication primitives.

Collaboration across research teams will shape model architectures and training recipes. You will prototype data, tensor, and pipeline parallelism strategies and help scale thousands of

Formación

• Bachelor’s degree or equivalent experience in CS, EE, stats, ML, physics, robotics, or similar.
• Understanding of deep learning frameworks (e.g., PyTorch, JAX) and their underlying system architectures.
• Thrive in a highly collaborative environment with cross-functional partners.
• Bias for action with initiative to ship across different stacks and teams.
• Strong engineering skills in floating-point numerics and distributed systems.

Responsabilidades

• Design and optimize distributed training infrastructure for large-scale LLMs, focusing on performance, stability, and reproducibility.
• Implement and evaluate low-precision numerics (BF16, MXFP8, NVFP4) to improve efficiency without sacrificing model quality.
• Develop kernels and communication primitives using hardware-level support for mixed arithmetic.
• Collaborate with research teams to co-design model architectures and training recipes for numeric formats.
• Prototype scaling strategies and integration of precision-adaptive computation and quantized communication.
• Contribute to orchestration and monitoring systems for thousands of distributed experiments.
• Publish and share learnings through internal docs or open-source libraries.

Educación

Bachelor’s degree or equivalent

Herramientas

PyTorch

JAX

PyTorch XLA

DeepSpeed

Megatron-LM

Descripción del empleo

Thinking Machines is hiring an infrastructure research engineer in San Francisco to design and build core systems for efficient large-scale model training. You will optimize distributed training stacks, implement low-precision numerics, and develop kernels with hardware-enabled communication primitives.

Collaboration across research teams will shape model architectures and training recipes. You will prototype data, tensor, and pipeline parallelism strategies and help scale thousands of