- Remote, Canada
- Remote
- Full-time · Senior
- 3+ years
Help train next generation of large language models by designing, running, and scaling training experiments on thousands of GPUs, and turning research ideas into production-grade models.
Responsibilities
- Design and run large-scale pre-training and fine-tuning experiments
- Study scaling laws to choose model size, data mix, and hyperparameters
- Improve training efficiency through parallelism strategies, mixed precision, and custom CUDA or Triton kernels
- Diagnose and fix training instabilities
- Build and improve data pipelines
Requirements
- PhD in computer science, mathematics, or physics
- Nice to have: Publications at top venues (NeurIPS, ICML, ICLR, ACL)
- Nice to have: Pre-training an LLM from scratch
- Nice to have: Supervised fine-tuning
- Nice to have: RLHF
