Member of Technical Staff Training Performance Engineer

Toronto-founded enterprise AI company building foundation models, agentic AI, enterprise search, and private deployment products.

Toronto, Canada
About Cohere

Cohere provides enterprise-focused AI models and end-to-end products, including North for agentic workplace AI, Compass for enterprise search, and models for generation and retrieval. It supports private, VPC, and on-premises deployment options.

View jobs by Cohere

Skills

About the Role

You will optimize language-model training performance, identify and remove bottlenecks, and design scalable training software. You will assess how architectural choices affect throughput and quality, write low-level GPU kernels, and experiment with training and data infrastructure.

Requirements

  • Software engineering
  • Python
  • JAX
  • PyTorch
  • XLA
  • MLIR
  • CUDA
  • Triton
  • GPU kernel development
  • Distributed training
  • Autoregressive model
  • Transformer

Responsibilities

  • Design and write high-performance scalable training software.
  • Evaluate architectural choices for their effects on training throughput and quality.
  • Write low-level CUDA and Triton kernels for accelerator performance.
  • Research, implement, and experiment with supercomputing and data infrastructure.

Benefits

  • Equity
  • Weekly lunch stipend of $75/£75 or equivalent local currency
  • Health and dental benefits
  • Mental health budget
  • RRSP matching, 401K, and Pension Scheme
  • 100% parental leave top-up for up to 6 months
  • Annual enrichment benefits
  • Education and learning stipend
  • 6 weeks of paid vacation
  • Budget for travel to other offices for remote employees
  • Annual company offsite
  • Daily lunch program for office employees
  • Snacks
  • Community and social events
  • Co-working benefit
  • $500 home office stipend