Scaling LLMs: Parameters or Thinking or Data?

​How do we truly scale the intelligence of large language models—by adding parameters, giving them more “thinking” time at inference, or exposing them to richer, task-specific data?

​​Join Prateek Jain — Principal Scientist & Director at Google DeepMind, and Research Lead on Gemini Model Design — as he unpacks the science and engineering behind scaling LLM quality. This talk will explore how different levers of scale (model size, test-time compute, and data curation) interact, and what that means for building the next generation of AI systems.

​​Here are some details:

​​📆 Date: 22nd Aug, 2025

​​⏰ Time: 5 to 6 PM onwards

​​📍 Location: Together Fund, Indiranagar, Bangalore

​​👤 About the speaker

​Prateek Jain is a Principal Scientist and Director at Google DeepMind, where he leads research on Gemini model design. With a distinguished career in machine learning and optimization, Prateek has worked at the frontier of scaling laws, model efficiency, and data-driven intelligence. His research and leadership continue to shape how cutting-edge AI models are designed, scaled, and deployed.