Google Dev: Local Models & Fine-Tuning

What can realistically run on a laptop or consumer GPU How quantization changes memory use, speed, and quality Fine-tuning versus prompting, retrieval, and tool use How parameter-efficient methods such as LoRA adapt a model without retraining every weight How to prepare a small, useful dataset and keep evaluation examples separate How to compare the base model with the tuned model so “better” means more than a good demo Where local inference improves privacy, cost, latency, offline access, or control—and where a hosted model is still the better choice Expect a technical talk, practical demos, honest tradeoffs, and time for questions. You do not need prior machine-learning experience, and you do not need a GPU to follow along. Plot twist: this is also the September OklahomAI meetup. Google Developer Group Tulsa, OklahomAI, and Techlahoma are bringing the communities together for one shared evening. Presented as part of Tulsa Tech Week 2026. Doors and conversation begin at 6:00 PM. The featured program starts at 6:15 PM. Presenter: Sam Carlton Live demos will run from Sam’s laptop. Bring your laptop if you want to follow along or test ideas during the session. Community Sponsor Black Dog Law Thank you to Black Dog Law for helping make accessible developer education possible in Tulsa.