AI Apps in a Flash: Ship to GPUs Without Docker — Dean Quiñanola, Runpod

Write code, build the image, push it, run it, find a bug, repeat. One loop can take 15 minutes. Runpod Flash cuts it out. Dean Quiñanola, staff engineer at Runpod, introduces Runpod Flash, which lets you run Python on cloud GPUs as if the GPU were local, with no Dockerfiles, image builds or registry pushes. He covers serverless queue and load-balanced endpoints, endpoints that call each other like local functions, network volumes with warm caching, and live serverless for instant iteration. In a live demo he checks which GPU he's on (an RTX 4090), evolves the same function into text generation and then image generation, installing dependencies on the fly, and deploys it to production with one command that packages and serves the app without Docker. He also shows off a hackathon project that fans experiments out across endpoints, and how to get started. In this talk: • Why the Docker build-push-run loop slows AI development • Live serverless: iterate on cloud GPUs like they're local • Endpoint-to-endpoint calls, network volumes and warm caching • Deploying to production with one command, no Docker SPEAKER Dean Quiñanola, Staff Software Engineer, Runpod GitHub: https://github.com/deanq LINKS Runpod Flash: https://runpod.io/flash Flash docs: https://docs.runpod.io/flash Flash (GitHub): https://github.com/runpod/flash Flash examples (GitHub): https://github.com/runpod/flash-examples CHAPTERS 0:00 Intro 0:27 The Docker loop 1:12 Why we built Flash 2:21 What Flash can do 2:41 Endpoint-to-endpoint calls 3:16 Live serverless 4:06 Demo: flash dev 5:40 Iterating: text generation 7:10 Fixing it live 7:59 Image generation 10:09 Deploying with flash deploy 11:13 Calling the deployed endpoint 12:18 A hackathon win 13:13 Getting started Recorded at the AI Engineer World's Fair 2026 in San Francisco. Subscribe for more talks from the engineers building with AI. AI Engineer: https://ai.engineer YouTube: https://www.youtube.com/@aiDotEngineer X: https://x.com/aiDotEngineer LinkedIn: https://www.linkedin.com/company/aidotengineer/ #Runpod #ServerlessGPU #AIEngineer