About the Role

As our Member of Technical Staff (Training Infrastructure), you’ll build and optimize our training infrastructure, designing efficient distributed training for supervised learning, RL, or new approaches that power our ambitious efforts in building efficient and robust models.

Your work will determine how fast we can train, how far we can push our models, and how quickly research becomes reality.

Responsibilities

• Build and optimize large-scale distributed training pipelines

• Develop scalable RL training frameworks for RLHF and policy optimization methods

• Experiment with new training techniques and integrate them into production

• Implement benchmarks for training throughput, convergence, and efficiency

• Optimize GPU/TPU utilization across hybrid cloud + on-prem clusters

• Collaborate closely with researchers to translate architectures into scalable training code

Qualifications

• Strong engineering skills in distributed ML training

• Proficiency in Python and performance-oriented languages (C++/Rust)

• Deep understanding of parallelism strategies — model, data, and pipeline

• Hands-on experience with DeepSpeed, FSDP, or similar frameworks

• Experience with hybrid cluster scheduling and orchestration (Kubernetes, Slurm)

What makes us interesting

• Small, elite team of ex-founders, researchers from top AI Labs, top CS grads, and engineers from top companies

• True ownership You will not be blocked by bureaucracy, shipping meaningful work within weeks rather than months

• Serious momentum We're well-funded by top investors, moving fast, and focused on execution

What we do

• Ship consumer products powered by cutting-edge AI research, and

• Build infrastructure that facilitates research and product, and

• Innovate cutting-edge research that will open up new consumer product forms

The Details

• Full-time, onsite role in Menlo Park

• Startup hours apply

• Generous salary, with additional benefits to be discussed during the hiring process