8th Neural Scaling Workshop @ NeurIPS

Hosted by Daria Soboleva & 4 others
Registration
2 Spots Remaining
Hurry up and register before the event fills up!
Approval Required
Your registration is subject to approval by the host.
Welcome! To join the event, please register below.
About Event

Come join us after NeurIPS for the 8th Scaling Workshop series that started in Oct 2021!

We provide a forum for discussing the challenges and advances in scaling foundation models. This workshop, co-organized by Cerebras Systems, MBZUAI and the CERC in Autonomous AI Lab led by Irina Rish at the Universite de Montreal and Mila - Quebec AI Institute will focus on both development and deployment aspects of large-scale models.

What to Expect

-> Frontier-scale training insights from leaders pushing the boundaries of pretraining, MoE architectures, and open foundation models.
-> Systems-level breakthroughs in distributed training and real-time inference from institutions like OpenAI, Snowflake, Vector Institute, and more.
-> Deep dives into model optimization — compression, ternary LLMs, hardware-aware design, and next-gen inference strategies.
-> Two high-profile panels featuring top thinkers shaping the geopolitics and economics of AI.
-> Evening receptions & networking with researchers, industry experts, and founders.

Program Highlights

Friday, December 5 from 5pm – Scaling Training & Distributed Systems

Vol Kyrylov (OpenAI) — GPT-OSS: 128 experts on a single GPU
Hector Liu (MBZUAI) — Pushing open foundation models to the limit
Marco Ciccone (Vector Institute) — Training LLMs across public supercomputers
Aurick Qiao (Snowflake) — Breaking the speed-cost tradeoff in LLM serving

Panel (7pm): “Sovereign AI: AI as Geopolitical Advantage” featuring: Natalia Vassilieva, Sara Hooker, Keunwoo Choi, Rio Yokota, Hrant Khachatrian 

Saturday, December 6 from 5pm – Efficient Inference & Model Optimization

Daria Soboleva (Cerebras) — MoE 101: Efficient training & serving
Junyang Lin (Qwen/Alibaba) — Deep dive into Qwen3
Eric Sather (Cerebras) — ML for high-performance inference
Irina Rish (UdeM/Mila/42.com) — Research perspectives on inference efficiency
Ayush Kaushal (Nolano AI/Mila) — Scaling laws & ternary LLM inference

Panel (7pm): “AI: Show Me the Money” Featuring: swyx, Dylan Patel, Irina Rish, Tri Dao


Saturday, December 6 at 8pm - Closing Reception 🍾🥳

Celebrate with speakers and participants!

Workshop website: https://sites.google.com/mila.quebec/8th-scaling-workshop/
Full Schedule: https://sites.google.com/mila.quebec/8th-scaling-workshop/schedule
Organizers:  Natalia Vassilieva, Daria Soboleva,  Karina Anichkina-Wolf,  Alexis Roger, Irina Rish
Moderators (sessions and panel):  Daria Soboleva

Location
Hard Rock Hotel San Diego
207 Fifth Ave, San Diego, CA 92101, USA
4 mins by walk from San Diego convention center