Hybride · Toronto
Architect and manage large-scale scraping and ingestion infrastructure for petabytes of image and video data. Design and optimize production ETL/ELT pipelines to prepare visual datasets for ML model training and evaluation.
Hybride · Toronto
Design and optimize high-throughput distributed training systems for multi-modal foundation models on large GPU clusters. Focus on resolving numerical instability, diagnosing hardware faults, and improving overall FLOPS utilization.
Hybride · Toronto
Design, deploy, and maintain large-scale bare-metal GPU clusters using Slurm and Kubernetes for distributed training. Optimize high-speed interconnects and distributed storage to ensure maximum GPU utilization and reliability.
Hybride · Toronto
Build and maintain high-throughput multimodal data pipelines for cleaning, filtering, and augmenting image and video datasets. Translate product goals into data strategies and coordinate large-scale labeling efforts to support foundation world models.