Back to news

AI · Video

Google DeepMind's Melvin Johnson on Post-Training Gemini 4 for Real Engineering Work

AI Engineer ·

At a glance

Google DeepMind's Melvin Johnson explains how Gemini 4 is post-trained for real engineering work. The talk covers harder training tasks, grading based on rubrics, and isolated test environments. It also looks at how Google's engineers' feedback helps models ship faster.

Key points

Why it matters

Developers and people curious about how AI coding agents are trained may find the talk useful.

Original source

Post-Training Gemini 4 for Real Engineering Work — Melvin Johnson, Google DeepMind

AI Engineer · Nov 18, 21:21 KST

Watch on YouTube ↗

AI summarized the video title and description. Check the original for accuracy. Copyright belongs to AI Engineer.

AI