From idea to production: Building AI apps with GDE Tomek Porozynski

Google Cloud · Advanced ·🛠️ AI Tools & Apps ·2h ago
How do you turn a complex AI idea into a production-ready app without getting lost in the hype? Join Aja Hammerly (Director DevX AI) and Tomek Porozynski (Google Developer Expert) live from Google Cloud Next '26 as they dissect the architecture of a Multi-Voice AI Audiobook generator. Tomek shares his journey of moving from a simple proof-of-concept to an open-source, scalable application using the latest Gemini APIs. The Architecture of an AI Audiobook: Character Extraction: Using Gemini to parse text, identify characters, and determine if the narrator is first or third person. Voice Assignment: Strategically selecting distinct voices for each character to ensure a cohesive listening experience. Serial vs. Parallel Processing: Tomek explains his "hybrid" workflow—parsing the story linearly while making multiple parallel calls to the Text-to-Speech (TTS) API to speed up production. Agent Skills & Up-to-Date Info: Why Tomek uses Agent Skills in the Gemini CLI to ensure his agents are aware of the latest model names and API interfaces that might have launched after the model's training cutoff. Pro-Tips for AI Developers: The 80/20 Rule of Planning: Tomek and Aja discuss why the "brainstorming phase"—having a long conversation with your agent about the tech stack before writing a single line of code—is the most critical step. Self-Correction & Troubleshooting: Don't give up on the first prompt. Tomek explains how to use Antigravity (IDE) to let agents spin up browsers, click through interfaces, and debug their own code. The "Slow Down" Method: In an era of AI "FOMO," Tomek’s advice is to start small. Use Google AI Studio to build the core logic, then scale it into Firebase for auth and Cloud Run for production-grade hosting. "AI technology evolves so fast that the model isn't always aware of the latest updates. That's where Skills come in—they give the model the guaranteed link to the most up-to-date resources." Get the Code: Check out Tomek's Multi-Voice Aud
Watch on YouTube ↗ (saves to browser)
Sign in to unlock AI tutor explanation · ⚡30

Related AI Lessons

Up next
New ChatGPT Workspace Agents are INSANE!
Julian Goldie SEO
Watch →