Friday, December 5th
4:30pm - 5:00 pm Check-in
5:00pm - 5:10pm - Daria Soboleva, Natalia Vassilieva, Irina Rish Intro and overview
5:10pm - 5:50pm - Session 1: Frontier model training
5:10pm - 5:30pm - Vol Kyrylov (OpenAI) gpt-oss: 128 experts for one GPU
5:30pm - 5:50pm - Hector Liu (MBZUAI) Pushing the limits of open foundation models
5:50pm - 6:30pm - Session 2: Distributed training & inference
5:50pm - 6:10pm - Marco Ciccone (Vector Institute) BOOM: Scaling LLMs training across public supercomputers
6:10pm - 6:30pm - Aurick Qiao (Snowflake) Arctic Inference: Breaking the speed-cost tradeoff in LLM serving
6:30pm - 7:00pm - Coffee break and Q&A
7:00pm - 8:00pm - Panel “Sovereign AI: The Case for Building Your Own”
Panelists: Natalia Vassilieva, Sara Hooker, Keunwoo Choi, Rio Yokota, Hrant Khachatrian, Preslav Nakov
Saturday, December 6th
4:30pm - 5:00 pm Check-in
5:00pm - 5:10pm - Daria Soboleva (Cerebras) Training and serving MoE models efficiently
5:10pm - 5:30pm - Junyang Lin (Qwen Team, Alibaba Group) Scaling model size and context length towards intelligence
5:30pm - 6:30pm - Session 3: Inference optimization
5:30pm - 5:50pm - Eric Sather (Cerebras) ML for High-Performance Inference
5:50pm - 6:10pm - Irina Rish (CERC-AAI UdeM Lab/Mila/42.com) Research perspectives on inference optimizations
6:10pm - 6:40pm - Session 4: Compression and optimization
6:10pm - 6:30pm - Ayush Kaushal (Nolano AI/Mila) Scaling laws & efficient inference for ternary LLMs
6:30pm - 7:00pm - Coffee break and Q&A
7:00pm - 8:00pm - Panel “AI: Show Me the Money”
Panelists: swyx, Dylan Patel, Irina Rish, Tri Dao, Hung Bui, Rahul Sengottuvelu
8pm - 9pm - Closing Reception 🎉🥳🍾