#144

Josh and Mike sit down with Kenny Warner, VP of Data Science and Engineering at meez, for a conversation that starts with Kenny's unlikely path into tech leaving college after his sophomore year to launch a cause-marketing startup and winds up deep inside the machinery of modern AI. Kenny breaks down what actually happens when a language model reads your data, explaining tokens, embeddings, attention, and inference in plain terms, and why you can't simply point a chatbot at a hundred-million-row database and expect good answers.

He and Josh dig into the real difference between building a model and fine-tuning one, why clean and trusted data is the true competitive advantage, and how smaller, purpose-built models can beat the giants at specific jobs. They also get into where this is all heading for restaurants, from turning recipes, costs, and margins into something a system can reason about to the rise of MCP and agentic tooling, with a detour through Nobu, a Miami chef tour, and a few fun facts along the way. It's a rare, jargon-free look under the hood of AI from someone who builds it every day.

Links and resources 📌

Visit meez: https://www.getmeez.com

Follow meez on Instagram: https://www.instagram.com/getmeez

Follow Josh on Instagram: https://www.instagram.com/joshlsharkey/?hl=en

Follow Josh on LinkedIn: https://www.linkedin.com/in/joshua-sharkey-406965b/

Follow Kenny Warner on LinkedIn: https://www.linkedin.com/in/kennywarner/

Visit Blanket: https://www.blanket.app/

Follow Michael on Instagram: @michaeljacober

Timestamps

08:04 Kenny's Path From College To Startup Founder

12:01 Raising Capital And Hitting Hard Times

15:01 Building A Model Versus Fine-Tuning One

16:11 What Tokens And Embeddings Really Are

18:45 How Attention Works In Language Models

26:01 Why Clean, Trusted Data Wins

29:05 The Case For Smaller, Purpose-Built Models

41:03 How AI Reads A Hundred Million Rows

43:21 MCP And The Rise Of Agentic Tools

50:01 What All This Means For Restaurants