Tinker: an API for Fine-Tuning & RL

Tinker by Thinking Machines (Mir Murati’s new AI lab) is a training + fine-tuning API that gives developers full control over the training loop—while it handles the heavy lifting of distributed infrastructure. In this live session, we’ll break down how Tinker works (LoRA, RL, VLM support, open-weights workflows) and what makes it different from other fine-tuning-as-a-service tools. Join us every Wednesday at 1pm EST for our live events. SUBSCRIBE and RING THE BELL to get notified! Join our community to start building, shipping, and sharing with us today! https://discord.gg/RzhvYvAwzA Get started learning end-to-end web dev for AI Engineers: https://maven.com/aimakerspace/ai-eng-onramp Or go all in with The AI Engineering Bootcamp! https://maven.com/aimakerspace/ai-eng-bootcamp Speakers: Dr. Greg, Co-Founder & CEO @ AI Makerspace https://www.linkedin.com/in/gregloughane The Wiz, Co-Founder & CTO @ AI Makerspace https://www.linkedin.com/in/csalexiuk/ 00:00 Fine-Tuning in 2026: Still Relevant? 06:00 What's New with Tinker and Fine-Tuning APIs? 11:00 The Cloud-First Training Workflow 17:00 How LoRA (Low-Rank Adaptation) Powers Modern Fine-Tuning 23:00 Adapters, Multi-Tenant Serving, and Model Efficiency 29:00 Tinker Demo: Fine-Tuning RL Models in Real Time 34:00 Hands-On with RL and LoRA: Guess the Number Game 42:00 Tinker in the AI Developer Tool Landscape 48:00 LoRA Deep Dive: Limitations, Capacity, and Research Insights 56:00 Future of Model Customization and Wrap-Up