Research Engineer, Infrastructure, Numerics (AI)

TL;DR

Research Engineer, Infrastructure, Numerics (AI): Designing and optimizing distributed training infrastructure for large-scale LLMs with an accent on performance, stability, and reproducibility. Focus on implementing low-precision numerics and developing efficient communication frameworks for scalable AI training.

Location: San Francisco, California

Compensation: $350,000 - $475,000 USD

Company

Thinking Machines Lab empowers humanity through advancing collaborative general intelligence.

What you will do

• Design and optimize distributed training infrastructure for large-scale LLMs.
• Implement and evaluate low-precision numerics to improve efficiency.
• Develop kernels and communication primitives for mixed and low-precision arithmetic.
• Collaborate with research teams on model architectures and training recipes.
• Prototype and benchmark scaling strategies for precision-adaptive computation.
• Contribute to internal orchestration and monitoring systems for distributed experiments.

Requirements

• Bachelor’s degree or equivalent experience in relevant fields.
• Understanding of deep learning frameworks like PyTorch and JAX.
• Strong engineering skills in complex codebases and distributed systems.
• Ability to thrive in a collaborative environment.

Nice to have

• Familiarity with distributed frameworks such as PyTorch/XLA and DeepSpeed.
• Experience with FP8, INT8 formats and their numerical trade-offs.
• Prior contributions to open-source deep learning infrastructure.
• Publications or projects related to numerical optimization.

Culture & Benefits

• Generous health, dental, and vision benefits.
• Unlimited PTO and paid parental leave.
• Relocation support as needed.