About Thomas Chong
Thomas Chong is an AI Research Engineer at Beever AI, a new frontier AI research lab by Votee AI, based in Toronto, Canada. He is a Google Developer Expert in AI, specializing in generative AI, NLP, multimodal large language models, and data visualization. Thomas has contributed to ACL 2025 with research on agentic LLMs for video explanations and benchmarks for Cantonese language understanding in LLMs.
Session
Live Model Fine-Tuning: Customizing Gemma 3 270M in Under 40 Minutes
VIP Demos (11:00 AM - 12:00 PM)
Witness the power of accessible AI as we fine-tune Google's new Gemma 3 270M model live on stage in under 40 minutes. This demo focuses on post-training customization, showcasing how a powerful pre-trained model can be rapidly adapted for a specialized task. To achieve this within the time limit, we will leverage the Unsloth library's remarkable speed and efficiency for Parameter-Efficient Fine-Tuning (PEFT) with LoRA, all within a standard Google Colab environment. This is a practical, step-by-step demonstration with no slides—just code and results.