How Developers Are Building Good AI with Gemma 4

Google announced the winners of the Gemma 4 Good Challenge, which called for the development of AI solutions to solve problems faced by people worldwide. Participants in this Kaggle competition leveraged key technologies such as LiteRT, Cactus, Ollama, llama.cpp, and Unsloth to achieve high performance even in resource-constrained environments.

The first-place winner, GEM-4, is a robotic assistant and control pipeline designed to physically assist the daily lives of the elderly and people with disabilities. The system uses the Gemma 4 31B model to label video learning clips and converts visual observations and language instructions into actual physical movements via a lightweight Gemma 4 E2B controller. Judges evaluated that this system, featuring an innovative closed-loop data engine and a VLA (Vision-Language-Action) architecture, connects AI reasoning directly to physical assistance.

The second-place winner, Trido, is a voice-controlled digital whiteboard designed specifically to help educators with physical disabilities. This tool focuses on managing interactive classroom environments, demonstrating another example of AI solutions addressing accessibility issues.