Optimize AI training and serving across Boson AI's 900-GPU cluster as a Toronto-based MScAC intern on a hybrid schedule.

This eight-month applied research internship suits University of Toronto MScAC students interested in efficient training and real-time inference. You may work with CUDA, Triton, SGLang, or vLLM while improving model speed, robustness, memory use, and serving capacity. The start date is May 2027, with earlier availability possible.

Key Responsibilities:

- Optimize numerical formats, kernels, checkpointing, and recovery
- Overlap communication with computation across multiple GPUs
- Improve quantization, decoding, and memory-layout performance
- Manage KV caches for concurrent users
- Scale streaming speech and video inference with demand

Requirements:

- Enrollment in University of Toronto MScAC program
- Robust coding skills in Python
- C++, CUDA, or Triton for systems projects
- Machine learning foundations with PyTorch or JAX experience
- Availability for the full eight-month hybrid internship

Apply systems optimization, GPU programming, and machine learning skills to Boson AI's large-scale training and inference infrastructure.

📌 GPU Systems Research Intern (Toronto)
🏢 Boson AI
📍 Toronto