What you need

3+ yrs software/ML engineering experience
Owned substantial system implementation to deployment
1+ yr training language/multimodal models
Built distributed GPU workloads with scheduling, containers, checkpointing
Production integration with customer data, APIs, cloud

What you'll do

Implement training algorithms and recipes
Build distributed training and inference systems
Develop data pipelines and evaluations
Create interfaces for experiments and model improvement
Deploy specialized models with customer data and workflows

Build the systems that turn research into useful models and products. We’re hiring research engineers across model training, ML infrastructure, full-stack product engineering, and customer deployment. Your focus will depend on your strengths and the problems you take ownership of.

You might implement training algorithms and recipes; build distributed training and inference systems; develop data pipelines and evaluations; create interfaces for running experiments and improving models; or work directly with customers to integrate their data and workflows and deploy specialized models. Across these areas, you’ll work closely with researchers and engineers and turn successful approaches into reliable, reusable capabilities.

Requirements

• Typically 3+ years of software or ML engineering experience, including relevant academic or industry research (preferred), or equivalent demonstrated work.

• Owned a substantial system from implementation through deployment and ongoing operation.

Role-specific requirements:

• Model training: typically 1+ year training language or multimodal models; hands-on PyTorch or JAX; experience building distributed training pipelines.

• ML infrastructure: built and operated distributed GPU workloads, including job scheduling, containers, checkpointing, failure recovery, and profiling training or inference performance.

• Customer deployment: owned a production integration involving customer data, APIs, and cloud infrastructure; debugged systems in customer environments and worked directly with technical customers.